NewOur benchmark results are out!read it

← The appendix · generated 2026-10-21

Infisical: every bot comment, every Heal defect

An open-source secrets-management platform.

46 PRs of its real history (2026-10-14 → 2026-10-16) replayed on a private copy, with CodeRabbit, cubic and Greptile on every PR at their default settings, Greptile's TREX on; the repo's own .greptile/config.json travels with the code and applies to Greptile. On the Heal side, a full run before the span (t1), one midway (t2) and one after it (t3).

100 defect-claim comments on 46 PRs, 94 distinct claims once the same point from two bots is merged; the 190 walkthroughs, summaries and "no issues" posts the bots also wrote are listed but not counted. 538 Heal defects shown, 21 of them on lines a replayed PR wrote, the rest on code from before the PRs tested. 16 security defects left out.

What ran, and when

The bots ran on their vendors' hosted services, on the settings above. The vendors publish no version number, so the posting dates below are the only identity of what reviewed these PRs.

  • coderabbit: 96 comments, posted between 2026-10-20 13:19 and 2026-10-21 00:24 UTC
  • cubic: 133 comments, posted between 2026-10-20 13:26 and 2026-10-21 00:24 UTC
  • greptile: 61 comments, posted between 2026-10-20 13:30 and 2026-10-21 00:31 UTC
Security defectsSecurity defects are left out of this appendix: the full run found 16 on this product, and its maintainers have not been told about them yet. Publishing an unfixed security defect before the people who can fix it have heard of it would put their users at risk, so those rows wait for the disclosure. They are counted in every figure of the benchmark page.

The comments are shown as the bots posted them, with no verdict on whether they hold; the same-issue judge only links a comment to the Heal defect it names. A Heal defect is a failing test on the running product. Download the comments and the defects as CSV.

This product's caveats
  • Judge truth: no human graded a sample on this product, so every judge-based number is the model's alone.
  • Blame names the last PR that touched a line, not the one that created the defect, so 'several PRs' and 'pre-span' are upper bounds on what a reviewer could have caught.
  • The defect ledger is cumulative over the runs: a defect filed by a later run may sit on lines an earlier PR of the span wrote.
  • The run after the span saw the code 14 PRs past the last replayed PR (the replay was stopped early by hand); a defect whose lines those later PRs wrote falls outside the window.
  • The bots' account settings are the vendors' defaults with Greptile's TREX on; the repo's own .greptile/config.json travels with the code and applies to Greptile.
  • Cost sums every metrics row of the workspace's runs, stopped attempts included; a run's summary card shows one segment only.
290 of 290 comments
PRBotWhereClaimsThe comment
#8635
2026-10-20 13:19
coderabbit
issue-comment
—summary
📝 Walkthrough The announcement modal now places navigation controls in a footer below the card stack. The footer includes Previous and Next/Done buttons, clickable announcement indicators, and a current-position display. The Storybook description reflects the updated controls. **Changes: Announcement modal** |Layer / File(s)|Summary| |:---|:---| |**Card stack and footer navigation** `frontend/src/components/announcements/AnnouncementModal.tsx`, `frontend/src/components/v3/platform/AnnouncementModal.stories.tsx`|The modal uses a flexible card-stack container and a bottom pager. The pager provides Previous and Next/Done buttons, clickable indicators, and a current-position display. Next cl
… the rest (2459 characters)oses the modal when the current announcement is the last one. The Storybook description documents the footer controls.| **Priority:** ⬇️ Low **Estimated code review effort:** 2 (Simple) | ~10 minutes **Change:** Feature **Suggested reviewers:** `andrewhuhh` Merge Risk: 🔵 Low · up to 043be On very narrow screens the Next or Done button may be partly pushed out of view in the announcement modal. This is a small layout issue that can be fixed after merge or just before it. Pre-merge checks: 5 | Check name | Status | Explanation | | :------------------------ | :--------------------------------------------------------------------------------------------: | :-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | Title check | | The title clearly and concisely describes the main UI change: making announcement modal navigation easier to discover. It follows the conventional commit format. | | Description check | | The description includes the required Context, Screenshots, Steps to verify the change, Type, and Checklist sections. It explains the previous and new behavior, documents verification steps, identifie… | | Docstring Coverage | | No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check. Docstring coverage is scoped to functions touched by this diff. Analyzed 0 functions across 2… | | Linked Issues check | | Check skipped because no linked issues were found for this pull request. | | Out of Scope Changes check | | Check skipped because no linked issues were found for this pull request. | Full details: Description check **Explanation** The description includes the required Context, Screenshots, Steps to verify the change, Type, and Checklist sections. It explains the previous and new behavior, documents verification steps, identifies the change as an improvement, and records testing and contribution-guide checks. Full details: Docstring Coverage **Explanation** No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check. Docstring coverage is scoped to functions touched by this diff. Analyzed 0 functions across 2 files. ✨ Finishing Touches 📝 Generate docstrings 🧪 Generate unit tests (beta) Comment `@the reviewer help` to get the list of available commands.
#8635
2026-10-20 13:23
coderabbit
inline · P2
frontend/src/components/announcements/AnnouncementModal.tsx:292defect-functional
**Keep the dot row within the footer.** The backend returns at most five announcements, but five fixed-width dots can still overflow a narrow footer. The pager does not wrap, and the navigation buttons cannot shrink. Constrain or horizontally scroll the dot row so `Next` or `Done` remains visible.
#8635
2026-10-20 13:23
coderabbit
review-body
—summary
**Actionable comments posted: 1**
#8635
2026-10-20 13:26
cubic
review-body
—summary
**No issues found** across 2 files View guided diff | Turn on auto-fix | Re-trigger the reviewer
#8635
2026-10-20 13:30
greptile
issue-comment
—summary
Confidence Score: 5/5 **[Low impact]** The change is safe to merge within the reviewed scope: navigation and keyboard-focus checks passed, and Done retains the existing close callback. Summary Replaces the side arrows with a navigation bar below the announcement card. - The announcement modal puts clear navigation controls below each multi-card announcement. Browser checks passed for button navigation, dot jumps, arrow keys, peeking previews, count synchronization, keyboard focus, Done closing, and the single-announcement layout. No actionable issues were found. The change is safe to merge based on the reviewed scope. Reviews (1) · Last reviewed commit: "improvement(ui): make announcem
… the rest (40 characters)ent modal..." · Reviewed by the reviewer
#8632
2026-10-20 13:32
coderabbit
issue-comment
—summary
No actionable comments were generated in the recent review. 🎉 📝 Walkthrough Select now supports controlled and uncontrolled open state. Overlay layers register Escape handlers and dispatch Escape events to the most recently registered popup. Select processes each event once and closes only when the event is not prevented. **Changes: Select Escape handling** |Layer / File(s)|Summary| |:---|:---| |**Overlay popup Escape dispatch** `frontend/src/components/v3/utils/overlay-layer.ts`, `frontend/src/components/v3/utils/overlay-layer.test.ts`|Overlay layers register optional Escape handlers, propagate popup registrations and dismissal events to parent layers, and invoke the most recently reg
… the rest (4438 characters)istered handler. Tests cover event cancellation, handler ordering, and cleanup.| |**Select open state and Escape integration** `frontend/src/components/v3/generic/Select/Select.tsx`|Select supports controlled and default open state, exposes state and Escape-event tracking through context, and registers popup dismissal handling. SelectContent invokes the optional Escape callback and closes only when the event is not prevented.| **Priority:** ⬇️ Low **Estimated code review effort:** 3 (Moderate) | ~20 minutes **Change:** Bug fix ### Sequence Diagram(s) ```mermaid sequenceDiagram participant guardDismiss participant SelectContent participant SelectOpenState guardDismiss->>SelectContent: Invoke the registered Escape handler SelectContent->>SelectContent: Process the keyboard event once SelectContent->>SelectOpenState: Close if the event is not prevented ``` **Suggested reviewers:** `andrewhuhh` Merge Risk: ⚪ Minimal · up to 185e9 No actionable Select dismissal issue remains identified. The failed Northflank build still needs diagnosis before the deployment is ready. Security architecture risk: 🔵 Low · up to 185e9 The change is confined to popup dismissal and preserves cancellation and parent-dismissal guards. No introduced security vulnerability was demonstrated. Remaining uncertainty concerns browser event ordering and mounted-component cleanup across nested overlays. **Retained concerns** No architecture-level concerns identified. Security review details **Security Blast Radius** - _inferred_ — The demonstrated exposure is browser-side dismissal behavior for a Select popup and its enclosing overlay ancestry. The inspected event path ends in UI callbacks and open-state updates; it does not demonstrate increased tenant, credential, service, or data-store authority. **Trust Boundaries and Controls** - _observed_ — Cancellation gates the registered Select close request, while event marking prevents the same Escape from also reaching an ancestor dismissal callback, including after synchronous unregistering. These are UI containment controls, not authentication or authorization boundaries. Pre-merge checks: 4 · 1 Failed checks: 1 warning | Check name | Status | Explanation | Resolution | | :---------------- | :----------------------------------------------------------------------------------------------: | :---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | :--------------------------------------------------------------------------------- | | Docstring Coverage | | Docstring coverage is 0.00% which is insufficient. The required threshold is 80.00%. Docstring coverage is scoped to functions touched by this diff. Analyzed 3 functions across 3 files. | Write docstrings for the functions missing them to satisfy the coverage threshold. | Passed checks: 4 passed | Check name | Status | Explanation | | :------------------------ | :--------------------------------------------------------------------------------------------: | :-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | Title check | | The title clearly and concisely describes the primary change: preventing Select Escape events from dismissing parent overlays. It follows the conventional commit format. | | Description check | | The description includes all required sections, explains the problem and solution, provides screenshots and verification steps, identifies the change type, and completes the checklist. It also clearly… | | Linked Issues check | | Check skipped because no linked issues were found for this pull request. | | Out of Scope Changes check | | Check skipped because no linked issues were found for this pull request. | Full details: Description check **Explanation** The description includes all required sections, explains the problem and solution, provides screenshots and verification steps, identifies the change type, and completes the checklist. It also clearly reports the remaining failed Northflank check and merge restriction. ✨ Finishing Touches: 💡 1 📝 Generate docstrings: 💡 🧪 Generate unit tests (beta) Comment `@the reviewer help` to get the list of available commands.
#8632
2026-10-20 13:43
greptile
issue-comment
—summary
Confidence Score: 5/5 **[Medium impact]** No blocking code defects were identified in the reviewed changes. Overall merge readiness remains blocked on the acknowledged Northflank preview-build failure, and PierreOucif has withheld merge authorization. Summary This PR keeps Select Escape events from closing enclosing overlays. Select shares its open state and cancelable Escape handler with the overlay layer, and popup registrations and consumed events propagate through enclosing layers. No actionable issues were found. On the current head, 40 browser cases passed across nested Dialog/Sheet compositions, cancellation, controlled refusal, callback delivery, and focus restoration. Twenty dis
… the rest (1137 characters)patched Escape synchronously when an option received focus; twenty used native keyboard input. The equivalent immediate-focus cases reproduced the enclosing-overlay dismissal on the base revision. All four focused bookkeeping tests passed, including original-event consumption through registration cleanup. These checks cover synthetic component compositions, not authenticated product/provider acceptance or every possible event interleaving. PierreOucif acknowledges the failed Northflank build and explicitly withholds merge authorization; its cause is not established by these changes. Overall merge readiness remains blocked on that acknowledged preview-build failure. No merge authorization is granted here. ### the check validation blocked Evidence upload could not be completed because this environment exposes no artifact-upload mechanism. Execution completed and evidence remains available locally, but no uploaded artifact references are claimed. - Escape closes an open Select before reaching its parent overlay. Reviews (1) · Last reviewed commit: "fix(ui): keep Select Escape from dismiss..." · Reviewed by the reviewer
#8632
2026-10-20 13:43
cubic
review-body
—summary
**No issues found** across 3 files View guided diff | Turn on auto-fix | Re-trigger the reviewer
#8462
2026-10-20 13:45
coderabbit
issue-comment
—summary
📝 Walkthrough The change adds PAM folder alerts for access-request events, emits those events within approval transactions, and adds resource-aware alert configuration to the frontend. It renames the PAM alerting entitlement while preserving legacy license-key support and existing Slack notification configurations. **Changes: PAM folder alerting** |Layer / File(s)|Summary| |:---|:---| |**Enterprise alerting entitlement compatibility** `backend/src/ee/services/license/*`, `backend/src/services/license-client/*`, `frontend/src/hooks/api/subscriptions/types.ts`, `frontend/src/components/license/UpgradeGate/*`, `tests/fakes/license/features.go`|The PAM feature name changes from Slack notific
… the rest (7338 characters)ations to Enterprise alerting. Legacy Slack entitlement keys still enable the renamed feature.| |**Transactional approval event emission** `backend/src/services/approval-policy/*`, `backend/src/ee/services/pam-access-request/pam-access-approval-resource.ts`, `backend/src/ee/services/pam-access-request/pam-access-request-events.ts`, `backend/src/server/routes/index.ts`, `backend/CLAUDE.md`|The approval policy service emits resource-built events within request, approval, rejection, and bypass transactions. PAM access-request events identify their folder and request targets.| |**PAM folder alert provider and lifecycle wiring** `backend/src/services/alert/providers/pam-folder-alert-*`, `backend/src/server/routes/index.ts`, `backend/src/ee/services/pam-folder/pam-folder-service.ts`, `backend/src/ee/services/pam/CLAUDE.md`|The provider resolves folder alert data, checks scope and permissions, builds alert payloads, and gates non-email channels by plan. Folder deletion removes associated alerts.| |**Resource-aware alert configuration** `frontend/src/hooks/api/alerts/types.ts`, `frontend/src/views/Alerts/*`, `frontend/src/pages/organization/IdentityDetailsByIDPage/components/IdentityAlertAction.tsx`, `frontend/src/pages/project/IdentityDetailsByIDPage/components/ProjectIdentityAlertAction.tsx`|The shared alert UI accepts resource-specific event types, email recipients for PAM folders, and channel paywall settings. Alert actions use a generic resource identifier.| |**PAM folder alert and legacy notification settings** `frontend/src/pages/pam/PamAccountsPage/components/FolderApprovalsTab.tsx`, `frontend/src/pages/pam/PamAccountsPage/components/FolderNotificationsSection.tsx`, `docs/documentation/platform/pam/access-requests/overview.mdx`|The PAM folder page adds alert controls and marks existing Slack notification configurations as read-only legacy settings. The documentation describes alert channels and events.| **Priority:** ➖ Normal **Estimated code review effort:** 4 (Complex) | ~45 minutes **Change:** Feature ### Sequence Diagram(s) ```mermaid sequenceDiagram participant ApprovalPolicyService participant PAMAccessApprovalResource participant EventOutboxService ApprovalPolicyService->>PAMAccessApprovalResource: Build event for approval lifecycle transition PAMAccessApprovalResource-->>ApprovalPolicyService: Return PAM folder event or null ApprovalPolicyService->>EventOutboxService: Emit event in the transaction ``` **Suggested reviewers:** `lb-vn`, `scott-ray-wilson` Merge Risk: 🔵 Low · up to 80283 Some users may be directed to connect a Slack integration even though legacy notification settings can no longer be added or edited. This is a bounded issue that does not block merging. Security architecture risk: 🔵 Low · up to 80283 The change adds notification destinations for privileged-access activity while retaining folder-level administration and organization isolation. No introduced security defect was confirmed, but recipient-delegation policy and end-to-end recovery guarantees remain incompletely established. **Retained concerns** No architecture-level concerns identified. Security review details **Security Blast Radius** - _inferred_ — The demonstrated new disclosure scope is metadata for access requests in a configured PAM folder, delivered to its selected notification destinations. Cross-organization and cross-project target selection is constrained by the request query and folder checks. Email recipients may include verified-domain addresses without folder membership, so notification delegation is broader than direct folder access. **Trust Boundaries and Controls** - _observed_ — Alert authorization first awaits PAM project membership and validates folder ownership. Reading requires folder audit-log authority or policy-management authority; writes require policy-management authority. Missing project membership throws before the resource-permission fallback can run. Built-in Folder Admin grants both actions, Auditor grants audit-log access, and Operator and Connector grant neither. - _observed_ — Directed delivery rechecks active, accepted organization membership and effective project membership for user recipients, including expanded group recipients. Explicit email addresses are checked against currently verified organization domains. These checks do not establish folder membership or reauthorize the alert creator at each send; the inspected design delegates destination selection through folder policy-management authority. **Resilience and Maintainability Implications** - _observed_ — Approval decisions use locked request state and emit lifecycle events inside the transaction that changes approval or grant state. Event insertion failures propagate through that transaction. The delivery engine tracks successful channels by event identifier and skips them on retry; partial delivery requests another attempt. This supports recovery but does not establish exactly-once external notification delivery. **Hardening Proposals** - _proposed_ — Make the PAM notification-delegation policy explicit: clarify whether folder administrators may intentionally send request metadata to project members and verified-domain addresses without folder access, and what revocation should mean for existing destinations. This is policy clarification, not a verified authorization finding. Pre-merge checks: 5 | Check name | Status | Explanation | | :------------------------ | :--------------------------------------------------------------------------------------------: | :-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | Title check | | The title clearly and concisely describes the main change: moving PAM approval notifications to the shared alerting module. It follows the conventional commit format. | | Description check | | The description includes the context, screenshots, change summary, type selection, and checklist. The Steps to verify section is empty, but the description is otherwise complete and directly related t… | | Docstring Coverage | | No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check. Docstring coverage is scoped to functions touched by this diff. Analyzed 0 functions across 3… | | Linked Issues check | | Check skipped because no linked issues were found for this pull request. | | Out of Scope Changes check | | Check skipped because no linked issues were found for this pull request. | Full details: Description check **Explanation** The description includes the context, screenshots, change summary, type selection, and checklist. The Steps to verify section is empty, but the description is otherwise complete and directly related to the pull request. Full details: Docstring Coverage **Explanation** No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check. Docstring coverage is scoped to functions touched by this diff. Analyzed 0 functions across 32 files. (3 skipped: 3 unsupported.) ✨ Finishing Touches 📝 Generate docstrings 🧪 Generate unit tests (beta) Comment `@the reviewer help` to get the list of available commands.
#8462
2026-10-20 13:53
coderabbit
review-body
—summary
> [!CAUTION] > Some comments are outside the diff and can’t be posted inline due to GitHub limitations. > > **⚠️ Outside diff range comments (1)** > > > 🟡 Minor · Remove the hint that tells users to connect a Slack… · FolderNotificationsSection.tsx:285-290 > > `frontend/src/pages/pam/PamAccountsPage/components/FolderNotificationsSection.tsx:285-290` > **🎯 Functional Correctness** | **🟡 Minor** | **⚡ Quick win** > > **Remove the hint that tells users to connect a Slack workflow integration.** > > This section is now read-only legacy UI. Users cannot add or edit Slack notification configs here, so connecting a new Slack workflow integration does not help. The hint still tells users to add o
… the rest (810 characters)ne under Organization Settings, Workflow Integrations, which is a dead end. > > The hint can also appear above working legacy configs. `integrationOptions` is empty when the viewer cannot list workflow integrations, and the code at Line 127-128 handles that case with the saved `integrationSlug` fallback. Remove the hint, or replace it with a pointer to the Alerts card. > > > Proposed fix > > ```diff > - {integrationOptions.length === 0 && ( > - > - No Slack workflow integrations are connected to this organization yet. Add one under > - Organization Settings, Workflow Integrations. > - > - )} > ``` > > > Based on learnings: "The older per-folder Slack configs (`pam_folder_notification_configs`) are legacy: they still post and can be removed, never added or edited." > > > > > > _Source: Learnings_ > >
#8462
2026-10-20 13:57
cubic
review-body
—summary
**No issues found** across 35 files View guided diff | Turn on auto-fix | Re-trigger the reviewer
#8462
2026-10-20 13:57
greptile
issue-comment
—summary
Confidence Score: 4/5 **[Medium impact]** The changes affect PAM approval alerts and the channels that receive them\. No blocking issue was established; the remaining concern is a temporary, non-blocking interruption to paid-channel configuration during rollout. Findings 1. **Paid channels appear locked** Summary Moves PAM approval alerts into the shared alert module while keeping existing Slack notifications active. - PAM folder alerts send updates when access requests change. - Folder approval settings now use the shared alert editor. - PAM enterprise alert channels use a broader plan entitlement. No blocking issue was established. Normalize or invalidate legacy plan caches to avoi
… the rest (112 characters)d this temporary interruption. Reviews (1) · Last reviewed commit: "address reviews" · Reviewed by the reviewer
#8462
2026-10-20 13:57
greptile
inline · P2
backend/src/services/alert/providers/pam-folder-alert-provider.ts:306defect-functional
**Paid channels appear locked** If pre-deployment cloud plan caches survive deployment, paid customers temporarily receive an upgrade error when adding Slack, webhook, or PagerDuty PAM alert channels. Those cached plans contain `pamSlackNotifications`, but this check reads only `pamEnterpriseAlerting`, and `getPlan` returns cached JSON without migrating the field. Access returns after the plan is refreshed; the cache lifetime is nominally 15 minutes, with jitter. This is a non-blocking rollout compatibility concern. Normalize the legacy field on cache reads, or invalidate/version the affected caches when deploying the rename. Artifacts **Executed PAM cache compatibility harness** - Loads
… the rest (1310 characters) repository TypeScript and exercises historical and PR-head entitlement behavior with stubbed infrastructure, making the reproduction inspectable and repeatable. **Executed command capture script** - Runs the before or after harness and records its command, working directory, exit code, stdout, and stderr, preserving observed execution evidence. **Legacy paid plan generated by historical projection** - The before run generated this fixture using the actual pre-rename projection, demonstrating the legacy cache contains the paid Slack grant without the renamed field. **Before rename: paid Slack entitlement accepted** - Executed historical cache reading, projection, reconciliation, and the original Slack entitlement guard through an HTTP adapter, showing the legacy paid cache is accepted. **PR head: legacy paid cache rejects non-email channels** - Executed the actual PR-head provider and error handler against legacy and refreshed plans, showing three erroneous 400 responses become 200 responses after fresh projection. **Source lines and unchanged application worktree** - Captured numbered source, historical cache-key comparison, cache-reference search, and Git status, identifying the compatibility gap and confirming no tracked application changes. Ran code and verified through TREX
#8636
2026-10-20 14:01
coderabbit
issue-comment
—summary
📝 Walkthrough The pull request adds typed analytics events and telemetry tracking across the Secrets overview page, its import and management workflows, and member activation surfaces. Tracking includes interactions, flow outcomes, resource counts, and invocation ownership. **Changes: Secrets Analytics** |Layer / File(s)|Summary| |:---|:---| |**Analytics contracts and activation tracking** `frontend/src/lib/analytics/events.ts`, `frontend/src/pages/secret-manager/OverviewPage/components/InviteMembersModal/InviteMembersModal.tsx`, `frontend/src/pages/secret-manager/OverviewPage/components/InviteMembersNudge/InviteMembersNudge.tsx`|Adds typed overview and activation events. The invite moda
… the rest (7062 characters)l and nudge capture activation events through organization-scoped analytics.| |**Telemetry lifecycle and overview context** `frontend/src/pages/secret-manager/OverviewPage/OverviewTelemetry.tsx`, `frontend/src/pages/secret-manager/OverviewPage/OverviewPage.tsx`|Adds count and outcome classification, flow lifecycle tracking, request observation, and React context. The overview records scope, viewed and loaded events, and popup flow ownership.| |**Overview operations and interactions** `frontend/src/pages/secret-manager/OverviewPage/OverviewPage.tsx`|Adds telemetry for folder and secret mutations, batch commits, filters, selection, search, navigation, sorting, pagination, and environment interactions.| |**Component actions and flow outcomes** `frontend/src/pages/secret-manager/OverviewPage/components/*`|Adds telemetry to create, copy, export, breadcrumb, row-edit, reveal, and bulk-selection actions. Tracked async flows use invocation ownership checks for completion and sheet closure.| |**Import flow tracking and ownership** `frontend/src/pages/secret-manager/OverviewPage/components/SecretDropzone/ImportSecretsModal.tsx`|Tracks import source, outcomes, counts, and cancellation. Centralized close handling checks invocation ownership before changing lock state or completing callbacks.| **Priority:** ➖ Normal **Estimated code review effort:** 4 (Complex) | ~45 minutes **Change:** Feature **Suggested reviewers:** `andrewhuhh` Merge Risk: 🟡 Moderate · up to 6078c Successful secret operations can leave forms or dialogs open under the identified scope and render transitions, and protected-branch commit analytics can report the wrong outcome. Resolve these issues before merging. Security architecture risk: 🟡 Moderate · up to 6078c The explicit event properties are carefully bounded, and the inspected changes preserve secret-operation permissions and approval handling. However, the analytics transport may attach browser URLs containing secret-folder paths despite those property restrictions. New visit and interaction events broaden that potential exposure. **Retained concerns** - **Medium · security · inferred:** The new events constrain custom properties but not SDK-generated browser metadata. Overview retains secretPath in its URL, and the existing transport forwards events to PostHog without a location-property scrubber. SDK URL enrichment can therefore carry secret-folder paths outside the intended allowlist. The transport condition predates this PR, but new visit and interaction captures broaden its exposure. Actual outgoing payloads were not available, so this is an inferred privacy concern rather than a verified disclosure. Security review details **Security Blast Radius** - _inferred_ — The privacy concern affects telemetry-enabled production browsers visiting V3 Overview pages or using the new interactions. Events are associated with organization and project identity. The supported exposure concerns browser-location metadata; inspected custom payloads do not carry secret values, clipboard text or file contents. **Security Findings and Attack Paths** - _inferred_ — A user-controlled folder path persists in the Overview query string. An ordinary visit or subsequent instrumented interaction reaches the SDK through the organization capture adapter. Default URL enrichment can then transmit that path despite the custom-property allowlist. Existing explicit captures shared this condition, but the new capture points increase its reach. Final SDK-enriched payloads remain unverified. **Trust Boundaries and Controls** - _observed_ — The existing transport requires production mode, enabled telemetry and an API key. Automatic page, click, exception and session-recording collection is explicitly disabled. These gates reduce collection scope but do not sanitize the new explicit captures. - _observed_ — The inspected Overview mutations preserve concrete project, environment and path arguments, folder permission filters, approval-response handling and context-keyed pending changes. The telemetry state machine does not grant mutation authority or replace those controls. **Resilience and Maintainability Implications** - _observed_ — The transport catches SDK capture exceptions, and the request observer returns original mutation results or rethrows original request errors. This keeps ordinary capture failures from replacing backend outcomes in the inspected execution path. **Hardening Proposals** - _proposed_ — Extend the privacy boundary to the final SDK-enriched payload: sanitize location and referrer properties or replace them with a fixed route, then validate serialized events for nested-folder visits and deep links. This would close the current proof gap without relying on custom-property selection alone. Pre-merge checks: 4 · 1 Failed checks: 1 warning | Check name | Status | Explanation | Resolution | | :---------------- | :----------------------------------------------------------------------------------------------: | :----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | :--------------------------------------------------------------------------------- | | Docstring Coverage | | Docstring coverage is 0.00% which is insufficient. The required threshold is 80.00%. Docstring coverage is scoped to functions touched by this diff. Analyzed 1 functions across 14 files. | Write docstrings for the functions missing them to satisfy the coverage threshold. | Passed checks: 4 passed | Check name | Status | Explanation | | :------------------------ | :--------------------------------------------------------------------------------------------: | :------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | Title check | | The title clearly and concisely describes the primary change: adding semantic telemetry for the Secrets Overview. It also follows the conventional commit format. | | Description check | | The description is complete and relevant. It includes context, screenshots, verification steps, change type, checklist status, testing details, scope limitations, and implementation constraints. | | Linked Issues check | | Check skipped because no linked issues were found for this pull request. | | Out of Scope Changes check | | Check skipped because no linked issues were found for this pull request. | ✨ Finishing Touches: 💡 1 📝 Generate docstrings: 💡 🧪 Generate unit tests (beta) ### Usage-based review receipt - Mode: Continue automatically - Reviewed files: 14 - Waived: $3.50 (charged $0.00) - View usage details > [!NOTE] > This review exceeded your plan’s limits and used usage-based reviews—free during trial, billed after paid activation unless disabled. **Manage usage-based reviews**. Comment `@the reviewer help` to get the list of available commands.
#8636
2026-10-20 14:10
cubic
inline · P2
frontend/src/pages/secret-manager/OverviewPage/components/SecretTableRow/SecretEditTableRow.tsx:1052defect-functional
A failed `refetchSharedValue()` returns before this flow starts, so failures while loading a secret for editing are absent from telemetry. Start the attempt before fetching and emit `failed` on this error path.
#8636
2026-10-20 14:10
cubic
inline · P2
frontend/src/pages/secret-manager/OverviewPage/components/SelectionPanel/SelectionPanel.tsx:196defect-functional
Bulk-move telemetry reports a zero or undercounted selection when rotations are selected because this bucket omits `selectedRotationCount`. Include the rotation count so the interaction reflects all selected items offered to the move flow. ```suggestion countBucket: overviewCountBucket( selectedFolderCount + selectedKeysCount + selectedRotationCount ) ```
#8636
2026-10-20 14:10
cubic
inline · P2
frontend/src/pages/secret-manager/OverviewPage/components/CopySecretsSheet/CopySecretsSheet.tsx:471defect-functional
Copy telemetry merges folder-creation responses into `responseCountBucket` and omits `folderResponseCountBucket`, so folder-only and secret-copy outcomes cannot be distinguished. Record the folder count separately in both success and partial-failure events.
#8636
2026-10-20 14:10
cubic
review-body
—summary
**3 issues found** across 14 files Reply with feedback, questions, or to request a fix.View guided diff | Turn on auto-fix | Re-trigger the reviewer
#8636
2026-10-20 14:12
coderabbit
inline · P2
frontend/src/pages/secret-manager/OverviewPage/components/CreateSecretForm/CreateSecretForm.tsx:184defect-functional
**Do not reassign `formOwner.current` during render.** Line 178 replaces `formOwner.current` during render when the telemetry instance changes. React can replay or discard a render. A discarded render can then reset `invocation` and make an in-flight submission look stale, so the form does not close. Do this reset in an effect keyed on `overviewTelemetry`. _Source: Linters/SAST tools_
#8636
2026-10-20 14:12
coderabbit
inline · P1
frontend/src/pages/secret-manager/OverviewPage/components/CreateSecretForm/CreateSecretForm.tsx:487defect-functional
**Do not require `attempt.isCurrent()` to close the form.** When the overview scope changes, `cancelOpenFlows()` deletes submitted flow states. A successful create that completes afterward fails `attempt.isCurrent()`, even when the invocation is still active. The form remains open with the submitted data. The default context is not a cause because its `isCurrent()` returns `true`. 🐛 Suggested fix ```diff --- "a/frontend/src/pages/secret-manager/OverviewPage/components/CreateSecretForm/CreateSecretForm.tsx" +++ "b/frontend/src/pages/secret-manager/OverviewPage/components/CreateSecretForm/CreateSecretForm.tsx" @@ -480,11 +480,10 @@ console.warn("failed to create secrets"); }
… the rest (580 characters) if ( allRequestsSucceeded && - attempt.isCurrent() && formOwner.current.invocation === invocation ) { onClose(); reset(); } ``` 📝 Committable suggestion > ‼️ **IMPORTANT** > Carefully review the code before committing. Ensure that it accurately replaces the highlighted code, contains no missing lines, and has no issues with indentation. Thoroughly test & benchmark the code to ensure it meets the requirements. ```suggestion if ( allRequestsSucceeded && formOwner.current.invocation === invocation ) { ```
#8636
2026-10-20 14:12
coderabbit
inline · P2
frontend/src/pages/secret-manager/OverviewPage/components/SecretTableRow/SecretEditTableRow.tsx:236defect-functional
**Do not write `editConfirmation.current` during render.** Line 236 writes the ref during render. `handleEditSecret` compares this ref after an `await`. A discarded render can therefore cause a wrong stale-confirmation decision, and the confirmation dialog stays open after a successful save. Sync the ref in an effect, or compare against popup state captured through a token set in `handlePopUpOpen`. _Source: Linters/SAST tools_
#8636
2026-10-20 14:12
coderabbit
inline · P2
frontend/src/pages/secret-manager/OverviewPage/OverviewPage.tsx:1361defect-functional
**Preserve popup ownership across discarded renders.** When `telemetryFlows` changes in a render that React discards, the render-time mutation of `popupOwners.current` can persist. A later render with the previous scope resets the owner map, while `popupOwnership` still contains the previous owner token. `isPopupCurrent` then rejects the close request. `usePopUp` stores open state separately and does not restore the owner. The `ImportSecretsSheet` close callback returns before calling `handlePopUpToggle`, so the popup can remain open. Moving the reset to `useEffect` is not sufficient. That effect can clear ownership after a committed scope change while the popup remains open. Store the ow
… the rest (154 characters)ner token and scope in committed popup state, and validate that record for each popup invocation. Do not use a `useMemo` snapshot as the ownership source.
#8636
2026-10-20 14:12
coderabbit
inline · P2
frontend/src/pages/secret-manager/OverviewPage/OverviewPage.tsx:2536defect-functional
**Use the protected-branch predicate for the commit flow phase.** `POST /api/v1/pit/batch/commit` returns only `{ message: "success" }`. It does not return the service result's `approvalId`. Therefore, protected-branch changes can be recorded as `response-received`. Proposed fix ```diff - await observeOverviewRequest(overviewTelemetry, flow, () => - createCommit({ - projectId, - environment: singleVisibleEnv.slug, - secretPath, - pendingChanges: changes, - message - }) - ); - const hasOnlyFolderChanges = changes.folders.length > 0 && changes.secrets.length === 0; const requiresApproval = isProtectedBranch &
… the rest (484 characters)& !hasOnlyFolderChanges; + const attempt = overviewTelemetry.startFlow(flow); + try { + await createCommit({ + projectId, + environment: singleVisibleEnv.slug, + secretPath, + pendingChanges: changes, + message + }); + } catch (error) { + attempt.flow({ phase: "failed" }); + throw error; + } + attempt.flow({ phase: requiresApproval ? "approval-requested" : "response-received" }); ```
#8636
2026-10-20 14:12
coderabbit
review-body
—summary
**Actionable comments posted: 5**
#8636
2026-10-20 14:19
greptile
issue-comment
—summary
Confidence Score: 4/5 **[Medium impact]** The telemetry changes affect the Secrets Manager Overview and its activation prompts\. No blocking issue was found; the abandonment telemetry gap and missing regression coverage are non-blocking follow-ups. Findings 1. **Page exits lose cancellation events** 2. **Editor safeguards lack regression tests** Summary Adds explicit organization-scoped telemetry for Secrets Overview. - Overview visits and usable loads now produce scoped telemetry events. - Overview controls now report navigation, filters, and batch choices. - Create and row-edit flows report progress without closing newer editors. - Imports report their source and observed results wit
… the rest (665 characters)hout taking over a replacement sheet. - Copy, export, and bulk actions report their observed outcomes. - Activation modal and card events now use organization-scoped capture. Two non-blocking follow-ups remain: leaving Overview with an unsubmitted create or import flow does not record cancellation, and the new editor ownership safeguards lack committed regression tests. Neither finding prevents merging. PierreOucif explicitly described retaining action-origin scope for anchored copy as intentional because the source and destination stay unchanged. Reviews (1) · Last reviewed commit: "feat: add semantic Secrets Overview tele..." · Reviewed by the reviewer
#8636
2026-10-20 14:19
greptile
inline · P2
frontend/src/pages/secret-manager/OverviewPage/OverviewPage.tsx:1194defect-functional
**Page exits lose cancellation events** Leaving Overview with an unsubmitted create or import sheet open does not record `canceled`. This effect calls `cancelOpenFlows()` when another scope renders, but has no unmount cleanup; neither sheet records cancellation when its parent unmounts. This is a non-blocking telemetry gap: abandonment counts depend on how users leave the page. Add lifecycle-safe unmount cleanup for open flows while preserving submitted requests and their eventual results. Artifacts **Evidence from the check** - The executed script extracts the relevant source and drives Chromium assertions, making the unmount evidence reproducible. **Command output from the check** -
… the rest (993 characters)The successful command captured exact extracted source, event payloads, lifecycle probes, and 18 passing checks, confirming missing unmount cancellation. **Recording of the check** - Chromium opens create and import flows and changes scope, showing that the existing effect emits canceled. **Scope-change control poster** - A browser screenshot captures the import scope-change control and observed cancellation, providing the recording's preview. **Recording of the check** - Chromium opens the same create and import flows and unmounts Overview, showing opened events without canceled despite confirmed cleanup. **Page-unmount reproduction poster** - A browser screenshot captures the replacement page and unmount lifecycle output with only opened telemetry, providing the recording's preview. **Command output from the check** - Executed unstaged and staged git diff checks returned exit code zero, confirming tracked source remained unchanged. Ran code and verified through TREX
#8636
2026-10-20 14:19
greptile
inline · P2
frontend/src/pages/secret-manager/OverviewPage/OverviewTelemetry.tsx:120tests
**Editor safeguards lack regression tests** The new flow state and request-ownership guards decide whether successful requests may close or reset editors, but no committed tests cover those safeguards. The existing Overview suite passes the same 59 tests in six unchanged files before and after this change. This is a non-blocking maintenance concern: later changes can silently break event counts or dismiss a replacement editor. Commit repeatable tests for retries, scope changes, replacement editors, and out-of-order responses; the temporary checks described in the PR cannot protect later changes. Note: If this suggestion doesn't match your team's coding style, reply to this and let me know.
… the rest (1591 characters) I'll remember it for next time! Artifacts **Evidence from the check** - Executed discovery and base/head comparison source. **Command output from the check** - Committed-test inventory, full Overview test inspection, unchanged-test comparisons, and source excerpts. **Command output from the check** - Actual frontend Vitest discovery output. **Command output from the check** - Base: six test files and 59 tests passed. **Command output from the check** - Head: six test files and 59 tests passed. **Command output from the check** - Captured execution of the discovery/comparison script. **Evidence from the check** - Executed temporary direct-helper test source; not committed regression coverage. **Evidence from the check** - Executed temporary probe configuration. **Evidence from the check** - Executed wrapper recording the probe command and output. **Command output from the check** - Initial probe harness import failure before tests executed. **Command output from the check** - Second probe harness mock-resolution failure before tests executed. **Command output from the check** - Corrected probe: three ownership tests passed. **Command output from the check** - Committed-test inventory, full Overview test inspection, unchanged-test comparisons, and source excerpts. **Command output from the check** - Base: six test files and 59 tests passed. **Command output from the check** - Head: six test files and 59 tests passed. **Command output from the check** - Corrected probe: three ownership tests passed. Ran code and verified through TREX
#8634
2026-10-20 14:20
coderabbit
issue-comment
—summary
📝 Walkthrough The cancellation feedback form now accepts an optional comment of up to 500 characters. It trims and validates the comment, displays privacy guidance, and includes non-empty comments with the cancellation reason in analytics. **Changes: Billing cancellation comments** |Layer / File(s)|Summary| |:---|:---| |**Analytics event contract** `frontend/src/lib/analytics/events.ts`, `ANALYTICS.md`|The billing cancellation event properties now allow an optional comment. The analytics guidance documents the 500-character limit, trimming, omission when empty, and restrictions on use and content.| |**Comment collection and submission** `frontend/src/pages/organization/BillingV2Page/comp
… the rest (5158 characters)onents/CancellationFeedbackDialog.tsx`|The dialog initializes and validates the optional comment, displays a character count, privacy guidance, and validation errors, and disables the textarea during submission. Analytics includes the comment only when it is non-empty.| **Priority:** ⬇️ Low **Estimated code review effort:** 2 (Simple) | ~10 minutes **Change:** Feature **Suggested reviewers:** `andrewhuhh` Merge Risk: 🔵 Low · up to f482a Screen-reader users do not receive the comment count as they type. The cancellation flow remains usable, but the count should be made accessible. Security architecture risk: 🟡 Moderate · up to f482a The optional comment creates a new path for potentially sensitive customer text to enter analytics. Length limits, privacy guidance, and submission only after successful cancellation constrain exposure, but they do not prevent sensitive content from being transmitted. The change does not introduce a new privileged operation or automatically collect secrets. **Retained concerns** - **Medium · security · inferred:** The new comment property permits customer-entered sensitive text to cross into organization-linked analytics. Trimming, a 500-character cap, and a warning constrain volume and guide users, but do not enforce the documented content restrictions. Sensitive disclosure is possible if a user includes such content; no actual disclosure is established. Security review details **Security Blast Radius** - _inferred_ — The directly exposed data is the cancelling user’s optional comment, associated with an organization, product, action, and reason in the configured analytics destination. The new code does not automatically read secret stores or confer additional billing authority. **Security Findings and Attack Paths** - _inferred_ — If a user pastes a credential or personal information into the comment and cancellation succeeds, the inspected path forwards that content to PostHog. This is a newly enabled disclosure path compared with the reason-only base, not evidence that sensitive content has actually been collected or exposed to unauthorized readers. **Trust Boundaries and Controls** - _observed_ — The UI warns against secrets, credentials, and personal data, applies a native length limit, and validates trimmed length. The documented exception is limited to this event property and prohibits sensitive values from code. Neither the inspected form nor its transport enforces content-level exclusion. **Resilience and Maintainability Implications** - _observed_ — The existing transition ordering is preserved: cancellation rejection leaves the dialog open without capture; success precedes capture and completion. Inputs and dismissal are disabled during submission, and the Telemetry capturer catches synchronous PostHog capture errors. - _inferred_ — Comment state is initialized on mount rather than reset when organization, product, or action props change. Ordinary dismissal and success close the flow, but whether a live organization change can preserve the mounted dialog remains unresolved; no cross-organization transfer is established. **Hardening Proposals** - _proposed_ — Treat this property as potentially sensitive feedback: establish restricted readership, retention, and deletion handling at the destination, and consider a dedicated feedback channel if those guarantees cannot be provided by analytics. Content screening could reduce accidental credential submission, but should not be represented as a complete safeguard. Pre-merge checks: 5 | Check name | Status | Explanation | | :------------------------ | :--------------------------------------------------------------------------------------------: | :-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | Title check | | The title clearly and concisely describes the main change: adding an optional cancellation comment to the Billing V2 cancellation event. | | Description check | | The description includes the required context, screenshots, verification steps, feature type, and checklist. It explains the UI, analytics behavior, privacy guidance, and testing scope. | | Docstring Coverage | | No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check. Docstring coverage is scoped to functions touched by this diff. Analyzed 0 functions across 2… | | Linked Issues check | | Check skipped because no linked issues were found for this pull request. | | Out of Scope Changes check | | Check skipped because no linked issues were found for this pull request. | Full details: Docstring Coverage **Explanation** No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check. Docstring coverage is scoped to functions touched by this diff. Analyzed 0 functions across 2 files. (1 skipped: 1 unsupported.) ✨ Finishing Touches 📝 Generate docstrings 🧪 Generate unit tests (beta) Comment `@the reviewer help` to get the list of available commands.
#8634
2026-10-20 14:21
cubic
review-body
—summary
**No issues found** across 3 files View guided diff | Turn on auto-fix | Re-trigger the reviewer
#8634
2026-10-20 14:24
coderabbit
inline · P2
frontend/src/pages/organization/BillingV2Page/components/CancellationFeedbackDialog.tsx:193defect-functional
**Make the changing character count accessible.** The count changes as the user types, but the textarea describes only the privacy hint. Screen-reader users do not receive the remaining-limit update from this markup. Give the count an ID, include it in `aria-describedby`, and use `aria-live="polite"` for its updates. (w3.org) Based on learnings: dynamic non-urgent status updates should use `aria-live="polite"` so screen readers can announce them. _Source: Learnings_
#8634
2026-10-20 14:24
coderabbit
review-body
—summary
**Actionable comments posted: 1**
#8634
2026-10-20 14:29
greptile
issue-comment
—summary
Confidence Score: 5/5 **[Medium impact]** Safe to merge within the reviewed scope; no actionable issues were found in the changed dialog or analytics property. Summary Adds optional cancellation comments to all three Billing V2 cancellation flows, with a 500-character limit, a live counter, and a sensitive-data warning. Trimmed, nonempty comments accompany the existing feedback event only after cancellation succeeds. Focused browser checks of the shared dialog passed: typing updated the counter, input stopped at 500 characters, blank and whitespace-only comments remained optional and omitted the comment property, and padded comments were trimmed. Pending cancellation disabled the textare
… the rest (762 characters)a and withheld feedback; failed cancellation preserved the input and sent no feedback. These checks used controlled cancellation callbacks and captured analytics calls locally, rather than a live billing account or PostHog destination. PierreOucif explicitly describes sending free-text feedback to PostHog for churn analysis as intentional. The PR adds the matching, narrowly scoped exception in ANALYTICS.md. PierreOucif also identifies a possible Slack destination as deferred and outside this PR. No actionable issues found. Safe to merge within the reviewed scope. - Billing V2 cancellation dialogs can send an optional comment with the reason. Reviews (1) · Last reviewed commit: "feat(billing): add optional cancellation..." · Reviewed by the reviewer
#8650
2026-10-20 14:32
coderabbit
issue-comment
—summary
📝 Walkthrough The Secret Scanning documentation is reorganized into guides for local and repository scans, provider setup, configuration, and findings. Navigation, redirects, and frontend links now point to the corresponding guides. **Changes: Secret Scanning documentation** |Layer / File(s)|Summary| |:---|:---| |**Navigation and scanning concepts** `docs/docs.json`, `docs/documentation/platform/secret-scanning/concepts/secret-scanning.mdx`, `docs/documentation/platform/secret-scanning/overview.mdx`, `docs/documentation/getting-started/choose-a-product/secret-scanning.mdx`, `docs/documentation/platform/secret-scanning/usage.mdx`|Navigation separates local and repository scanning and adds
… the rest (4179 characters) links to dedicated guides. The overview and concepts pages describe scanning approaches. The former usage page is removed, and its URL redirects to the repository scanning guide.| |**Local scanning workflows and rules** `docs/documentation/platform/secret-scanning/scan-locally.mdx`, `docs/documentation/platform/secret-scanning/pre-commit-hook.mdx`, `docs/documentation/platform/secret-scanning/ignore-findings.mdx`, `docs/documentation/platform/secret-scanning/local-configuration.mdx`, `docs/snippets/documentation/platform/secret-scanning/*`|The local guides cover CLI scans, pre-commit hooks, ignored findings, and local configuration. Snippets document rule fields, allowlists, and inline-ignore syntax.| |**Repository scans and provider setup** `docs/documentation/platform/secret-scanning/scan-repositories.mdx`, `docs/documentation/platform/secret-scanning/github.mdx`, `docs/documentation/platform/secret-scanning/gitlab.mdx`, `docs/documentation/platform/secret-scanning/bitbucket.mdx`|The repository guide describes data sources, resources, historical and realtime scans, and Auto-Scan. Provider guides describe connection requirements, setup, repository selection, and scan history.| |**Repository configuration and findings** `docs/documentation/platform/secret-scanning/configuration.mdx`, `docs/documentation/platform/secret-scanning/findings.mdx`, `frontend/src/pages/secret-scanning/SecretScanningFindingsPage/components/SecretScanningFindingsSection.tsx`, `frontend/src/pages/secret-scanning/SettingsPage/components/ProjectScanningConfigTab/SecretScanningConfigForm.tsx`|The configuration guide documents repository scan rules and allowlists. The findings guide covers finding details, statuses, updates, deduplication, and notifications. Frontend documentation links now point to the findings and configuration guides.| **Priority:** ⬇️ Low **Estimated code review effort:** 3 (Moderate) | ~25 minutes **Change:** Other **Suggested reviewers:** `adkah`, `mathnogueira` Merge Risk: 🔵 Low · up to 7c333 Readers could misunderstand licensing, remediation, or how to configure local scans. These are localized documentation corrections; the reorganization is otherwise mergeable with those corrections tracked. Pre-merge checks: 5 | Check name | Status | Explanation | | :------------------------ | :--------------------------------------------------------------------------------------------: | :-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | Title check | | The title clearly describes the Secret Scanning documentation overhaul and follows the conventional commit format. | | Description check | | The description explains the documentation overhaul, identifies the change type, and includes the relevant checklist confirmations. The required steps-to-verify section is omitted, but the description… | | Docstring Coverage | | No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check. Docstring coverage is scoped to functions touched by this diff. Analyzed 0 functions across 2… | | Linked Issues check | | Check skipped because no linked issues were found for this pull request. | | Out of Scope Changes check | | Check skipped because no linked issues were found for this pull request. | Full details: Description check **Explanation** The description explains the documentation overhaul, identifies the change type, and includes the relevant checklist confirmations. The required steps-to-verify section is omitted, but the description is otherwise sufficiently complete for this documentation change. Full details: Docstring Coverage **Explanation** No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check. Docstring coverage is scoped to functions touched by this diff. Analyzed 0 functions across 2 files. (18 skipped: 18 unsupported.) ✨ Finishing Touches 📝 Generate docstrings 🧪 Generate unit tests (beta) Comment `@the reviewer help` to get the list of available commands.
#8650
2026-10-20 14:38
coderabbit
inline · P2
docs/documentation/platform/secret-scanning/concepts/secret-scanning.mdx:11docs
**Do not promise automatic detection and revocation.** A repository scans each new push only when Auto-Scan is enabled. A scan records a finding; the user must revoke an exposed credential. The current sentence can lead readers to assume that detection and remediation happen without those steps. State both conditions explicitly. (infisical.com)
#8650
2026-10-20 14:38
coderabbit
inline · P2
docs/documentation/platform/secret-scanning/concepts/secret-scanning.mdx:17docs
**Link to both configuration workflows.** This paragraph covers local and repository scanning, but its link opens only the repository configuration guide. A local CLI user who follows it will reach dashboard instructions instead of the `.infisical-scan.toml` instructions. Add a separate link to the local configuration guide. (infisical.com)
#8650
2026-10-20 14:38
coderabbit
inline · P2
docs/documentation/platform/secret-scanning/overview.mdx:3docs
**Include uncommitted code in the overview.** The description limits secret scanning to committed code. The CLI also scans changes before a commit, which is a principal workflow on this page. Describe scanning as checking code for exposed secrets before or after commit. The concepts page has the same narrow definition at Line 9. (infisical.com)
#8650
2026-10-20 14:38
coderabbit
inline · P2
docs/documentation/platform/secret-scanning/scan-repositories.mdx:12docs
**State that repository scanning requires the paid add-on.** The pricing page says secret scanning is a paid add-on to Enterprise, not part of the base plan. Line 12 can lead a reader to purchase Enterprise and expect repository scanning to be available. State that the Enterprise plan also requires the secret-scanning add-on. (infisical.com)
#8650
2026-10-20 14:38
coderabbit
review-body
—summary
**Actionable comments posted: 4**
#8650
2026-10-20 14:42
cubic
inline · P2
docs/documentation/platform/secret-scanning/pre-commit-hook.mdx:19docs
Both examples print detected secret values, which can persist in terminal scrollback or commit/CI logs. Add `--redact` to both commands.
#8650
2026-10-20 14:42
cubic
inline · P2
docs/documentation/platform/secret-scanning/overview.mdx:8docs
This limits Secret Scanning to already-committed credentials, but the CLI also scans directories and uncommitted changes. Revise both this sentence and the page description to include pre-commit scans.
#8650
2026-10-20 14:42
cubic
inline · P2
docs/documentation/platform/secret-scanning/concepts/secret-scanning.mdx:11docs
This overstates what scanning does: scans run only when configured or invoked, and Infisical records findings rather than revoking credentials. Describe detection as conditional and revocation as a response teams must take. ```suggestion As teams grow, it gets easy for a secret to end up in version control, a CI/CD pipeline, or a shared file. Anyone who can read that code can then use the secret to reach production systems, third-party services, or internal APIs. A secret scanner can detect leaks when you run it or configure automatic scans, giving your team a chance to revoke them before someone else finds them. ```
#8650
2026-10-20 14:42
cubic
review-body
—summary
**6 issues found** across 24 files Reply with feedback, questions, or to request a fix.View guided diff | Turn on auto-fix | Re-trigger the reviewer
#8650
2026-10-20 14:42
cubic
inline · P2
docs/documentation/platform/secret-scanning/configuration.mdx:15docs
This claim is false for `.infisical-scan.toml`: without a saved configuration, the repository scan omits `-c`, and the CLI automatically loads that file from the checkout. Correct the list or disable config auto-discovery before saying repository scans ignore it.
#8650
2026-10-20 14:42
cubic
inline · P2
docs/documentation/platform/secret-scanning/scan-repositories.mdx:12docs
State that repository scanning requires the Secret Scanning add-on in addition to Enterprise; the current wording implies the base Enterprise plan includes it. ```suggestion Scanning repositories requires Infisical's Enterprise plan and the Secret Scanning add-on. If you're self-hosting Infisical, contact sales@infisical.com to purchase an enterprise license. ```
#8650
2026-10-20 14:42
cubic
inline · P3
docs/documentation/platform/secret-scanning/concepts/secret-scanning.mdx:17docs
Link to both configuration guides here; local CLI users currently land on repository settings instead of the `.infisical-scan.toml` instructions. ```suggestion Infisical's scanner includes rules for hundreds of services. You can add your own rules and skip values you've decided are safe in the [repository scanning configuration](/documentation/platform/secret-scanning/configuration) or the [local scanning configuration](/documentation/platform/secret-scanning/local-configuration). ```
#8650
2026-10-20 14:45
greptile
issue-comment
—summary
Confidence Score: 4/5 **[Low impact]** The changes stay within Secret Scanning documentation and its help links\. No blocking issue was established; the documentation can merge with a non-blocking compatibility concern affecting previously shared section links. Findings 1. **Preserve existing section links** Summary Reorganizes Secret Scanning documentation into local scanning and connected repository workflows. - The Secret Scanning docs guide readers to separate local and repository scanning paths. - Local scanning guides explain how to scan code and handle findings before a push. - Repository scanning guides explain how data sources scan connected code. - A dedicated findings guide
… the rest (1149 characters)explains what secret findings contain and how teams review them. - Repository configuration guides show how to set detection rules and skip safe values. The new navigation destinations, updated product documentation links, and shared MDX imports passed focused checks. One non-blocking compatibility issue remains in `docs/docs.json`: existing findings and configuration section links on the old usage page no longer reach their corresponding guidance. No blocking issue was established. The email-service prerequisite remains consistent with the implementation: the server checks SMTP configuration before creating in-app notifications or sending email. ### the check validation blocked Native Mintlify visual verification was blocked because the pinned mint@4.2.13 CLI rejects pre-existing logo.href and contextual assistant/mcp configuration fields on both the base and head revisions. Representative pages and bookmark behavior were exercised in a fallback MDX renderer instead; deployed Mintlify redirect behavior was not observed. Reviews (1) · Last reviewed commit: "docs(secret-scanning): overhaul section ..." · Reviewed by the reviewer
#8650
2026-10-20 14:45
greptile
inline · P2
docs/docs.json:4876defect-functional
**Preserve existing section links** Existing `/documentation/platform/secret-scanning/usage#findings` and `/documentation/platform/secret-scanning/usage#configuration` links now redirect to `scan-repositories`, which contains neither section anchor. Readers following previously shared links or bookmarks therefore lose their section destination instead of reaching the new findings or configuration guide. This is non-blocking, but retaining compatibility sections with links to the moved guides would avoid that extra search. Note: If this suggestion doesn't match your team's coding style, reply to this and let me know. I'll remember it for next time! Artifacts **Evidence from the check** -
… the rest (2048 characters) The authored harness reads exact Git objects and executes MDX with explicit snippet resolution and substitute components; its scope and rendering limitations are documented in the source. **Evidence from the check** - The authored Playwright script starts the fallback servers and asserts representative page interactions and bookmark behavior; the browser evidence is reproducible. **Evidence from the check** - The authored runner executes integrity checks, native preview attempts, and Chromium capture while saving command metadata and actual output; logs are generated from executed commands. **Command output from the check** - The focused checker validates original navigation, links, assets, frontend hrefs, and MDX execution; all base checks pass. **Command output from the check** - The focused checker validates new routes, both frontend hrefs, nine imports, and all scoped MDX execution; direct destinations pass while legacy fragment loss is recorded. **Command output from the check** - The browser capture script records successful page loads, nested snippet visibility, imported-anchor navigation, and legacy bookmark outcomes; both runs finish without page errors. **Recording of the check** - Chromium visits the original configuration, findings, local guide, and legacy bookmarks using base-commit MDX; both old usage anchors exist. **Recording of the check** - Chromium exercises the new configuration and findings pages, expanded nested snippets, and local configuration; new content renders but legacy usage anchors are absent. **Original repository configuration in the fallback MDX renderer** - A Chromium screenshot captures the original configuration section with its example expanded; this is fallback rendering rather than the Mintlify UI. **Migrated configuration with nested snippet properties expanded** - A Chromium screenshot captures the new configuration page after expanding shared rule properties; nested allowlist content is present in fallback rendering. Ran code and verified through TREX
#8582
2026-10-20 14:47
coderabbit
issue-comment
—summary
No actionable comments were generated in the recent review. 🎉 📝 Walkthrough Webhook alert channels now support bearer, basic, and custom-header authentication, plus optional custom headers. Saved credentials are redacted and can be reused only while the webhook URL remains unchanged. Frontend forms, backend delivery, tests, and documentation cover these settings. **Changes: Webhook Channel Configuration and Delivery** |Layer / File(s)|Summary| |:---|:---| |**Validate and persist webhook configuration** `backend/src/services/alert/alert-channel-types.ts`, `backend/src/services/alert/alert-channel-config-fns.ts`, `backend/src/services/alert/channels/alert-channel-registry.ts`, `backend/s
… the rest (6644 characters)rc/services/alert/alert-channel-service.ts`, `backend/src/services/alert/alert-channel-service.test.ts`, `backend/src/services/alert/channels/alert-channels.test.ts`, `backend/e2e-test/routes/v1/alerts.spec.ts`, `backend/e2e-test/testUtils/alerts.ts`|Webhook validation now covers authentication modes, custom-header constraints, and saved-credential reuse. Channel creation and updates store parsed configuration. Tests cover validation, credential redaction and updates, header trimming, and URL-change rejection.| |**Build authenticated webhook requests** `backend/src/services/alert/channels/alert-channel-webhook-fns.ts`, `backend/src/services/alert/alert-channel-test-service.ts`, `backend/src/services/alert/alert-channel-test-service.test.ts`, `backend/src/services/alert/channels/alert-channels.test.ts`|Webhook requests combine configured headers, authentication headers, and JSON content type. Channel testing parses incoming and stored configurations before delivery.| |**Model webhook form data** `frontend/src/hooks/api/alerts/types.ts`, `frontend/src/hooks/api/alerts/utils.ts`, `frontend/src/pages/cert-manager/components/CertificateAlerts/types.ts`, `frontend/src/pages/secret-manager/OverviewPage/components/SecretTableRow/SecretReminderForm.tsx`, `frontend/src/views/Alerts/ChannelsField.tsx`|Form schemas and conversion helpers handle authentication credentials, saved-secret state, signing-secret state, and custom headers. Alert form integrations use shared channel conversion and initialize header collections.| |**Render webhook controls and guidance** `frontend/src/views/Alerts/ChannelCard.tsx`, `frontend/src/views/Alerts/ChannelSecretInput.tsx`, `frontend/src/views/Alerts/WebhookAuthenticationField.tsx`, `frontend/src/views/Alerts/WebhookHeadersField.tsx`, `frontend/src/views/Alerts/WebhookSigningSecretField.tsx`, `docs/documentation/platform/identities/machine-identities.mdx`, `docs/documentation/platform/pki/applications/alerting/webhook-alerts.mdx`|The alert form renders authentication, signing-secret, and custom-header controls. Documentation describes authentication options, credential reuse, and header constraints.| **Priority:** ⬇️ Low **Estimated code review effort:** 4 (Complex) | ~45 minutes **Change:** Feature ### Sequence Diagram(s) ```mermaid sequenceDiagram participant sendWebhookNotification participant triggerWebhook participant safeRequestPost sendWebhookNotification->>triggerWebhook: Pass parsed config and payload triggerWebhook->>safeRequestPost: POST with configured and authentication headers ``` **Suggested reviewers:** `thiago-as`, `carlosmonastyrski`, `lobocv` Merge Risk: ⚪ Minimal · up to 8ea36 No actionable merge-blocking issue remains; webhook credentials are not forwarded through redirects. Security architecture risk: 🔵 Low · up to 8ea36 Encrypted storage, destination binding, permission checks, and outbound request controls provide meaningful protection. No new vulnerability was established, but credential protection in failed-request logs and deployment-specific network exceptions were not fully confirmed. **Retained concerns** No architecture-level concerns identified. Security review details **Security Blast Radius** - _inferred_ — Outbound requests gain the authority conferred by the configured endpoint credentials; their downstream privileges cannot be determined from this repository. Saved-channel test reuse remains subject to organization, project, channel-type, and owning-alert edit checks. **Security Findings and Attack Paths** - _observed_ — The supplied redirect-exfiltration candidate is countered by the actual POST dispatch, which forces maxRedirects to zero. A redirect response does not automatically forward the authentication headers to another destination. No verified Security finding remains in the supplied assessment. **Trust Boundaries and Controls** - _observed_ — The shared outbound boundary rejects private addresses by default and pins validated DNS results. Existing development and internal-network exceptions remain configuration-dependent. Header validation protects transport and signing headers, while delivery constructs generated authentication and content-type headers after custom headers. **Resilience and Maintainability Implications** - _observed_ — Webhook requests retain bounded timeout and cancellation signals. Production dispatch catches delivery failures and records failure status, but also passes the full exception to logging. Whether the existing logger redacts every new credential-bearing header remains unresolved; this is a coverage limitation, not a verified disclosure. Pre-merge checks: 5 | Check name | Status | Explanation | | :------------------------ | :--------------------------------------------------------------------------------------------: | :-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | Description check | | The description covers the problem, implemented behavior, UI changes, verification steps, feature type, and documentation impact. The checklist items remain unchecked, and the verification steps conta… | | Title check | | The title clearly and concisely describes the main change: adding authentication and custom headers to webhook channels. It follows the conventional commit format. | | Docstring Coverage | | No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check. Docstring coverage is scoped to functions touched by this diff. Analyzed 0 functions across 2… | | Linked Issues check | | Check skipped because no linked issues were found for this pull request. | | Out of Scope Changes check | | Check skipped because no linked issues were found for this pull request. | Full details: Description check **Explanation** The description covers the problem, implemented behavior, UI changes, verification steps, feature type, and documentation impact. The checklist items remain unchecked, and the verification steps contain a duplicate number, but the required content is otherwise complete. Full details: Docstring Coverage **Explanation** No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check. Docstring coverage is scoped to functions touched by this diff. Analyzed 0 functions across 21 files. (2 skipped: 2 unsupported.) ✨ Finishing Touches 📝 Generate docstrings 🧪 Generate unit tests (beta) Comment `@the reviewer help` to get the list of available commands.
#8582
2026-10-20 14:52
greptile
issue-comment
—summary
Confidence Score: 4/5 **[Medium impact]** Webhook alert delivery changes reach several alert setup flows and their configured endpoints\. Do not merge until the authentication tabs satisfy the repository’s shared-component styling requirement. Findings 1. **Authentication tabs override shared spacing** Summary Adds authentication and custom headers to webhook alert channels. - Webhook alerts can use authentication and custom headers. Reviews (1) · Last reviewed commit: "improve naming" · Reviewed by the reviewer
#8582
2026-10-20 14:52
greptile
inline · P1
frontend/src/views/Alerts/WebhookAuthenticationField.tsx:66quality
**Authentication tabs override shared spacing** `WebhookAuthenticationField` replaces the shared `TabsContent` spacing with `mt-0` here and on line 77. This removes the component’s default `mt-3` spacing from the authentication panels and violates the repository directive that new UI must not override v3 design-system component defaults. Remove both margin overrides, or introduce a shared component variant if compact tabs are required, before merging. **Rule Used:** New UI must not override or deviate from the default styling of v3 design-system components (frontend/src/components/v3). Follow DESIGN.md. Flag: legacy v2 color tokens (e.g. mineshaft-*, bunker-*) instead of semantic tokens li
… the rest (164 characters)ke bg-org or text-dange... (source) Note: If this suggestion doesn't match your team's coding style, reply to this and let me know. I'll remember it for next time!
#8582
2026-10-20 14:59
cubic
inline · P2
frontend/src/hooks/api/alerts/types.ts:258defect-functional
This schema accepts more than 20 headers, but the backend rejects arrays above 20, so 21+ entries pass client validation and fail on submit. Add a `.max(20, ...)` constraint here. ```suggestion headers: z .array( z.object({ key: z.string().trim().min(1, "Header name is required"), value: z.string().min(1, "Header value is required") }) ) .max(20, "A webhook channel can have at most 20 headers") .default([]), ```
#8582
2026-10-20 14:59
cubic
review-body
—summary
**4 issues found** across 23 files View guided diff | Turn on auto-fix | Re-trigger the reviewer
#8582
2026-10-20 14:59
cubic
inline · P2
frontend/src/views/Alerts/WebhookHeadersField.tsx:65defect-functional
The Add Header button remains enabled after 20 entries, but the server rejects more than 20 headers, leaving the channel unsavable. Disable or hide the button at the limit. ```suggestion <Button type="button" variant="outline" size="xs" className="self-start" isDisabled={fields.length >= 20} onClick={() => append({ key: "", value: "" })} ```
#8582
2026-10-20 14:59
cubic
inline · P2
backend/src/services/alert/channels/alert-channel-registry.ts:33defect-security
Custom headers can carry API keys, but channel details return their values verbatim because `headers` is not redacted. Redact custom header values and expose only names and has-value flags.
#8582
2026-10-20 14:59
cubic
inline · P2
docs/documentation/platform/pki/applications/alerting/webhook-alerts.mdx:63docs
This says all credentials are hidden and blank fields preserve them, but Basic Auth usernames and API Key header names remain visible and must stay populated on updates. Clarify that only secret values are hidden and reused when left blank. ```suggestion Infisical hides saved secret values; leave those fields blank to keep them. Basic Auth usernames and API Key header names remain visible when you edit the channel. If you change the URL, re-enter the secret value, because Infisical only sends it to the URL it was saved with. ```
#8526
2026-10-20 15:01
coderabbit
issue-comment
—summary
No actionable comments were generated in the recent review. 🎉 📝 Walkthrough The change adds persisted audit-log stream failure state, transactional delivery-failure events, and organization-wide alert support. The frontend adds alert configuration for audit-log streams and displays failure status and error details. **Changes: Audit-log stream delivery alerts** |Layer / File(s)|Summary| |:---|:---| |**Persist stream health** `backend/src/db/migrations/*`, `backend/src/db/schemas/audit-log-streams.ts`, `backend/src/ee/services/audit-log-stream/audit-log-stream-service.ts`|The stream schema and migration add nullable failure fields. Saving credentials clears the failure fields.| |**Update
… the rest (7410 characters) delivery and failure state** `backend/src/ee/services/audit-log-stream-outbox/*`, `backend/e2e-test/audit-log-stream-outbox.integration.spec.ts`, `backend/e2e-test/fakes/*`, `backend/e2e-test/setup/reset-fakes.ts`, `backend/vitest.e2e.config.mts`|The outbox processes delivery results and stale claims in transactions, tracks stream failures, and emits failure events. Tests and e2e fakes cover retries, drops, recovery, and failure transitions.| |**Route failure events to organization alerts** `backend/src/services/alert/providers/audit-log-stream-alert-*`, `backend/src/server/routes/index.ts`, `backend/e2e-test/routes/v1/audit-log-stream-failure-alert.spec.ts`, `backend/e2e-test/testUtils/license.ts`, `backend/CLAUDE.md`|The alert provider handles delivery-failure events and resolves affected streams. Route registration and an end-to-end test cover alert delivery to an organization recipient.| |**Configure alerts and display stream health** `frontend/src/hooks/api/alerts/types.ts`, `frontend/src/hooks/api/auditLogStreams/types/providers/root-provider.ts`, `frontend/src/pages/organization/AuditLogsPage/components/AuditLogStreamTab/*`, `frontend/src/pages/organization/IdentityDetailsByIDPage/components/IdentityAlertAction.tsx`, `frontend/src/pages/project/IdentityDetailsByIDPage/components/ProjectIdentityAlertAction.tsx`, `frontend/src/views/Alerts/*`|The frontend adds audit-log stream alert types and setup, supports resource-less alerts, and displays failure details in stream rows. Identity alert actions now provide their resource type.| **Priority:** ➖ Normal **Estimated code review effort:** 4 (Complex) | ~45 minutes **Change:** Feature ### Sequence Diagram(s) ```mermaid sequenceDiagram participant OutboxService participant OutboxDAL participant EventOutbox participant AlertProvider participant AlertDAL OutboxService->>OutboxDAL: Commit delivery result and stream health OutboxService->>EventOutbox: Emit delivery-failed event in transaction EventOutbox->>AlertProvider: Dispatch delivery-failed event AlertProvider->>AlertDAL: Resolve stream IDs and providers ``` **Suggested reviewers:** `thiago-as`, `scott-ray-wilson`, `calvin486` Merge Risk: ⚪ Minimal · up to a4bd4 This change persists audit-log stream health, raises one organization-wide alert when a stream begins dropping logs, and shows a failing badge in the UI. No defects were confirmed in failure tracking, recovery, alert routing, or the shared alert form changes. Existing identity and PAM alert screens still pass their resource types. The change appears ready to merge after normal checks. Security architecture risk: 🟡 Moderate · up to a4bd4 The change improves visibility into audit-log delivery failures and preserves organization-level access controls. However, a failure while recording the notification can still allow delivery work to be permanently dropped while the stream appears healthy. The new failure-alert guarantee therefore remains incomplete. **Retained concerns** - **Medium · reliability · observed:** The new notification contract permits an exhausted delivery to be dropped without persisted failure health or a durable alert event. Both delivery and stale recovery delete exhausted rows before calling flagStreamFailing. If health mutation or event emission fails, that helper rolls back only its savepoint, logs the error, and returns successfully, allowing the deletion to commit. A healthy stream can consequently remain displayed as healthy after losing delivery work; if no further drops occur, the documented next-drop retry cannot repair the missing notification. This compromises audit-delivery monitoring rather than introducing the preexisting drop policy. Security review details **Security Blast Radius** - _inferred_ — A failure of one stream can trigger the organization's configured stream alerts, rather than only a stream-specific notification. The inspected target lookup constrains stream resolution to the alert organization; the new provider does not establish cross-organization authority. **Trust Boundaries and Controls** - _observed_ — Creation invokes server-side provider permission and scope assertions using the actor's organization. Listing checks provider Read permission and filters by that organization. The provider maps these operations to Settings permissions; the frontend Settings gate is not the sole authorization control. **Resilience and Maintainability Implications** - _observed_ — The rollback test proves that explicitly aborting a transaction can restore dropped rows, health, and events together. It calls the DAL and emitter directly, however, so it does not establish that the production helper propagates an emission failure. That helper deliberately catches the failure and preserves the outer deletion. **Hardening Proposals** - _proposed_ — Define an explicit externally safe error-content contract, using classified messages or redaction before persistence and notification. Length limiting alone does not establish confidentiality, although the inspected evidence did not prove a credential-disclosure path. Pre-merge checks: 4 · 1 Failed checks: 1 warning | Check name | Status | Explanation | Resolution | | :---------------- | :----------------------------------------------------------------------------------------------: | :-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | :--------------------------------------------------------------------------------- | | Docstring Coverage | | Docstring coverage is 0.00% which is insufficient. The required threshold is 80.00%. Docstring coverage is scoped to functions touched by this diff. Analyzed 2 functions across 29 files. (1 skipped: 1… | Write docstrings for the functions missing them to satisfy the coverage threshold. | Passed checks: 4 passed | Check name | Status | Explanation | | :------------------------ | :--------------------------------------------------------------------------------------------: | :----------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | Title check | | The title is concise, uses conventional commit format, and clearly identifies the primary change: alerts for audit-log streams that start dropping logs. | | Description check | | The description covers the context, behavior changes, verification steps, screenshots, feature type, and most checklist items. It is relevant and sufficiently complete. | | Linked Issues check | | Check skipped because no linked issues were found for this pull request. | | Out of Scope Changes check | | Check skipped because no linked issues were found for this pull request. | Full details: Docstring Coverage **Explanation** Docstring coverage is 0.00% which is insufficient. The required threshold is 80.00%. Docstring coverage is scoped to functions touched by this diff. Analyzed 2 functions across 29 files. (1 skipped: 1 unsupported.) **Resolution** Write docstrings for the functions missing them to satisfy the coverage threshold. ✨ Finishing Touches: 💡 1 📝 Generate docstrings: 💡 🧪 Generate unit tests (beta) Comment `@the reviewer help` to get the list of available commands.
#8526
2026-10-20 15:13
cubic
inline · P2
backend/src/ee/services/audit-log-stream-outbox/audit-log-stream-outbox-service.test.ts:609tests
This only checks that the sweep runs more than once, so a regression that raises the batch cap still passes. Assert the expected cap to protect the bounded-work guarantee. ```suggestion expect(recoverStaleClaims.mock.calls).toHaveLength(10); ```
#8526
2026-10-20 15:13
cubic
inline · P2
backend/src/ee/services/audit-log-stream/audit-log-stream-service.ts:204defect-functional
This clears the failure state on filter- or mode-only updates that resubmit unchanged credentials; the validation ping can succeed while actual batch delivery still fails. Clear health only after credentials actually change or a real delivery succeeds, or the failing indicator disappears and the next drop emits another alert.
#8526
2026-10-20 15:13
cubic
review-body
—summary
**3 issues found** across 30 files View guided diff | Turn on auto-fix | Re-trigger the reviewer
#8526
2026-10-20 15:13
cubic
inline · P2
backend/src/ee/services/audit-log-stream-outbox/audit-log-stream-outbox-service.ts:421defect-concurrency
A clean batch can erase a newer failure when stream drains overlap: its send may succeed before another drain drops rows, but its transaction can clear the flag afterward. Gate the clear on a success newer than `failingSince` or serialize result commits per stream.
#8526
2026-10-20 15:14
greptile
issue-comment
—summary
Confidence Score: 4/5 **[Medium impact]** The changes affect audit log stream delivery and alerts within an organization\. Changes are required before merging because an alert-insert failure can permanently remove audit logs without saving the configured notification or failing status. Findings 1. **Dropped logs go unreported** Summary Adds organization-wide alerts when an external audit log stream starts dropping logs. - Audit log streams record delivery failures and alert on dropped rows. - Organizations can get alerts when an audit log stream starts failing. - Audit Logs settings now let people manage stream alerts and see stream health. Do not merge until dropping the logs also gu
… the rest (170 characters)arantees a saved notification or durable notification retry. Reviews (1) · Last reviewed commit: "Merge remote-tracking branch 'origin/mai..." · Reviewed by the reviewer
#8526
2026-10-20 15:14
greptile
inline · P1
backend/src/ee/services/audit-log-stream-outbox/audit-log-stream-outbox-service.ts:247defect-data
**Dropped logs go unreported** If the alert-event insert fails, `flagStreamFailing` rolls back its savepoint and catches the error, but the outer transaction still commits deletion of the exhausted audit-log rows. The stream retains no failing status and no alert is queued. With no further logs, subsequent drains and stale-claim sweeps cannot retry the notification, so an organization can permanently lose logs without receiving its configured alert. Before merging, let this error roll back the drop too, or save a durable notification retry before committing the deletion. Artifacts **Evidence from the check** - Authored harness loads the actual historical and head services, injects a dat
… the rest (1644 characters)abase-side failure, and asserts persisted state—the executed test source is included. **Evidence from the check** - Runner initializes an isolated PostgreSQL cluster, captures both checks with commands and exit codes, and stops the server—the verification is reproducible. **Command output from the check** - Checked installed database commands and provisioning options—PostgreSQL was initially absent, but Debian package tooling was available. **Command output from the check** - Downloaded and extracted Debian PostgreSQL packages into the workspace and checked linked libraries—a real server could be run without installing a system service. **Command output from the check** - Initialized and started a fresh user-owned cluster with TCP disabled—the test used an isolated real database. **Command output from the check** - Executed the historical service with a failing event insert, then recovered the database and swept the retained stale claim—the row survived rollback and an alert was eventually persisted. **Command output from the check** - Executed the head service under the same insertion failure and checked recovery—the deletion committed while health and alert records were absent, with no retry after recovery. **Command output from the check** - Captured the exact head, candidate source lines, comparison revision, runtime versions, and clean tracked diff—the check targeted the requested candidate without production edits. **Command output from the check** - Stopped the isolated PostgreSQL server after both checks—the verification left no test database server running. Ran code and verified through TREX
#8638
2026-10-20 15:16
coderabbit
issue-comment
—summary
📝 Walkthrough This pull request adds OpenAPI operation identifiers and ACS documentation. It also extends test clients and fake DNS, adds an in-process SAML identity provider, and introduces integration tests for SAML configuration, login, redirects, assertion validation, and enforcement. **Changes: SAML integration test support and coverage** |Layer / File(s)|Summary| |:---|:---| |**Document SAML and email-domain API operations** `backend/src/ee/routes/v1/email-domain-router.ts`, `backend/src/ee/routes/v1/saml-router.ts`, `backend/src/lib/api-docs/constants.ts`, `tests/clients/api/oapi-codegen.yaml`|Adds OpenAPI operation identifiers for email-domain and SAML routes, documents ACS param
… the rest (7069 characters)eters, and updates the generated-client operation allowlist.| |**Configure test clients and tenant access** `tests/harness/infisical/client.go`, `tests/harness/infisical/client_test.go`, `tests/harness/infisical/infisical.go`, `tests/harness/tenant.go`, `tests/docs/client.md`|Adds configurable client options, redirect control, anonymous tenant clients, and a method that returns the test server URL. Documents redirect and cookie handling.| |**Support TXT records in fake DNS** `tests/infra/fakenet/*`, `tests/cmd/fakenet/main.go`, `tests/docs/fakes.md`|Adds TXT record storage, DNS responses, admin endpoints, and a publishing helper with cleanup. Updates the DNS server call, adds DNS tests, and documents TXT record setup.| |**Build and verify the SAML identity-provider fake** `tests/fakes/samlidp/*`, `tests/go.mod`|Adds a fake SAML IdP that parses redirect requests and generates configurable signed responses. Adds tests for request parsing, signing, tampering, and non-success responses, plus XML-signing dependencies.| |**Set up SAML integration fixtures and configuration tests** `tests/suites/sso/saml/helpers_test.go`, `tests/suites/sso/saml/main_test.go`, `tests/suites/sso/saml/config_test.go`|Adds shared SAML test setup and helpers, suite initialization, and tests for creating, reading, and updating SAML configurations.| |**Cover SAML login and redirect flows** `tests/suites/sso/saml/login_test.go`, `tests/suites/sso/saml/redirect_test.go`|Adds tests for SAML signup and login, account linking, identity attributes, redirect routing, callback ports, and inactive or unknown configurations.| |**Cover SAML validation and enforcement** `tests/suites/sso/saml/assertion_test.go`, `tests/suites/sso/saml/provider_test.go`, `tests/suites/sso/saml/enforcement_test.go`|Adds tests for invalid assertions and email domains, provider-specific signing and audience behavior, and organization SAML enforcement.| **Priority:** ➖ Normal **Estimated code review effort:** 4 (Complex) | ~60 minutes **Change:** Other **Suggested reviewers:** `akhilmhdh` Merge Risk: 🔵 Low · up to 920d4 Production behavior changes only through API documentation metadata and a new, documentation-oriented body schema on the SAML login callback. The rest adds SAML end-to-end tests and test infrastructure. One small follow-up remains: add generous field limits to the new callback schema to meet the project's input-validation standard. Request size is already capped, so this is safe to merge with that follow-up. Security architecture risk: 🔵 Low · up to 920d4 The inspected SAML changes retain the existing authentication flow. Most additions support integration testing, and no introduced security weakness was established. Some interface behavior, network exposure, and test-state cleanup details remain unverified, preventing a conclusive minimal-risk assessment. **Retained concerns** No architecture-level concerns identified. Security review details **Security Blast Radius** - _inferred_ — The demonstrated downstream scope is SAML integration testing and fake-network DNS state. The response configuration has multiple test dependents, but high fanout does not itself establish production exposure. External reachability of the TXT admin interface and isolation among concurrent test domains remain unresolved. **Trust Boundaries and Controls** - _observed_ — The inspected base-to-head SAML comparison retains passport authentication in preValidation, before schema validation, and retains configuration-route authentication hooks. Operation identifiers and the ACS body declaration do not replace those controls or create a new unauthenticated route. Exact unknown-field behavior was not independently verified. Pre-merge checks: 4 · 1 Failed checks: 1 warning | Check name | Status | Explanation | Resolution | | :---------------- | :----------------------------------------------------------------------------------------------: | :-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | :--------------------------------------------------------------------------------- | | Docstring Coverage | | Docstring coverage is 21.74% which is insufficient. The required threshold is 80.00%. Docstring coverage is scoped to functions touched by this diff. Analyzed 46 functions across 23 files. (4 skipped:… | Write docstrings for the functions missing them to satisfy the coverage threshold. | Passed checks: 4 passed | Check name | Status | Explanation | | :------------------------ | :--------------------------------------------------------------------------------------------: | :-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | Title check | | The title clearly identifies the primary change: adding SAML API end-to-end test coverage. It follows the conventional commit format, although the wording could be more precise about the broader test … | | Description check | | The description includes the required Context, Screenshots, Steps to verify, Type, and Checklist sections. It provides a basic summary and verification command, and the generated summary supplies addi… | | Linked Issues check | | Check skipped because no linked issues were found for this pull request. | | Out of Scope Changes check | | Check skipped because no linked issues were found for this pull request. | Full details: Title check **Explanation** The title clearly identifies the primary change: adding SAML API end-to-end test coverage. It follows the conventional commit format, although the wording could be more precise about the broader test coverage. Full details: Description check **Explanation** The description includes the required Context, Screenshots, Steps to verify, Type, and Checklist sections. It provides a basic summary and verification command, and the generated summary supplies additional implementation context. Several checklist items remain unchecked, but the description is mostly complete. Full details: Docstring Coverage **Explanation** Docstring coverage is 21.74% which is insufficient. The required threshold is 80.00%. Docstring coverage is scoped to functions touched by this diff. Analyzed 46 functions across 23 files. (4 skipped: 4 unsupported.) **Resolution** Write docstrings for the functions missing them to satisfy the coverage threshold. ✨ Finishing Touches: 💡 1 📝 Generate docstrings: 💡 🧪 Generate unit tests (beta) ### Usage-based review receipt - Mode: Continue automatically - Reviewed files: 27 - Waived: $6.75 (charged $0.00) - View usage details > [!NOTE] > This review exceeded your plan’s limits and used usage-based reviews—free during trial, billed after paid activation unless disabled. **Manage usage-based reviews**. Comment `@the reviewer help` to get the list of available commands.
#8638
2026-10-20 15:23
coderabbit
inline · P2
backend/src/ee/routes/v1/saml-router.ts:362hardening
**Add explicit ACS field bounds, but do not add a second pre-passport validator.** The new `SAMLResponse` and `RelayState` properties lack the `.max()` bounds required by the backend validation guide. Add limits based on the supported SAML payload contract. Do not choose an arbitrary smaller limit: Zod validation runs after `passport.authenticate`, so it may reject a response that passport already accepted. Fastify already applies its default 1 MiB `bodyLimit` while parsing the form body, before `preValidation`. The requested pre-passport total-size protection therefore already exists.
#8638
2026-10-20 15:23
coderabbit
review-body
—summary
**Actionable comments posted: 1**
#8638
2026-10-20 15:26
cubic
inline · P1
tests/suites/sso/saml/config_test.go:191defect-security
This makes a certificate-only update silently disable organization-wide SSO enforcement. Preserve enforcement when rotating the IdP certificate and assert it remains enabled here, or password logins regain access.
#8638
2026-10-20 15:26
cubic
inline · P2
tests/docs/fakes.md:113docs
The literal `value` will not verify an Infisical email domain; the verifier requires `infisical-domain-verification=<VerificationCode>` at the record name returned by the API. Show the required value and record name in this example.
#8638
2026-10-20 15:26
cubic
inline · P2
tests/suites/sso/saml/helpers_test.go:144defect-functional
Azure SP-initiated responses expect `spn:<issuer>`, but this default uses `SITE_URL`, so `spLogin` fails unless each Azure caller overrides `Audience`. Choose the default based on the configured provider and login flow.
#8638
2026-10-20 15:26
cubic
inline · P2
backend/src/ee/routes/v1/saml-router.ts:361design
Require `SAMLResponse` in this schema; the ACS can only authenticate with an IdP assertion, but OpenAPI currently lets generated clients omit it and send requests Passport rejects. ```suggestion SAMLResponse: z.string().describe(SamlSso.ACS.SAMLResponse), ```
#8638
2026-10-20 15:26
cubic
review-body
—summary
**4 issues found** across 29 files **Tip**: the reviewer can generate docs of your entire codebase and keep them up to date. Try it here.View guided diff | Turn on auto-fix | Re-trigger the reviewer
#8638
2026-10-20 15:30
greptile
issue-comment
—summary
Confidence Score: 5/5 **[Low impact]** No blocking findings were identified; the checked support behavior passed, while complete application-backed SAML behavior remains unverified. Summary Adds API end-to-end tests for SAML login, signup, redirects, provider signatures, configuration, and enforcement. - SAML tests now cover login, setup, redirects, and enforcement. - Tests can use a fake SAML identity provider to drive login responses. - fakenet can publish and answer test DNS TXT records. - The API client and test harness can call and inspect SAML endpoints. The container-free identity-provider, DNS, and HTTP-client support tests passed. Local checks also verified TXT publication and
… the rest (850 characters)resolution, long-value chunking, isolated record cleanup, composed request headers, ACS form encoding, and redirect inspection. No blocking defect was identified in the checked paths. PierreOucif explicitly preserves the existing enforcement reset on configuration updates and intentionally leaves the ACS strings unbounded. Neither is reported as a new defect. ### the check validation blocked The selected application-backed SAML signup, returning-user login, and invalid-signature tests could not run because their 180-second timeout expired while the backend Docker image was still building. No selected test began and no product endpoint was reached. Complete SAML signup, session creation, and invalid-signature rejection remain unverified. Reviews (1) · Last reviewed commit: "feat: addressed the reviewer review" · Reviewed by the reviewer
#8592
2026-10-20 15:33
coderabbit
issue-comment
—summary
No actionable comments were generated in the recent review. 🎉 📝 Walkthrough The certificate authority installation flow now uses v3 dialog and form components. App connection forms support both sheet and dialog layouts, with the sheet layout remaining the default. **Changes: Certificate authority installation dialogs** |Layer / File(s)|Summary| |:---|:---| |**Add dialog layout to app connection forms** `frontend/src/pages/organization/AppConnections/AppConnectionsPage/components/AppConnectionForm/AppConnectionForm.tsx`, `.../AppConnectionFormContext.tsx`, `.../AppConnectionFormFooter.tsx`|`AppConnectionForm` accepts an optional `"sheet"` or `"dialog"` layout and defaults to `"sheet"`.
… the rest (5085 characters)Create and update forms adjust their layout classes based on the setting. The footer renders `DialogFooter` for dialog layout and retains the sheet footer for sheet layout.| |**Migrate certificate authority installation dialogs** `frontend/src/pages/cert-manager/CertificateAuthoritiesPage/components/CaInstallCertModal/*`, `AGENTS.md`|The installation modal and ADCS, external, internal, and Venafi forms now use v3 dialog and form components. Connection creation passes `layout="dialog"` to `AppConnectionForm`. The development guidance now directs v2 `Modal` parents hosting v3 overlays to migrate to v3 `Dialog`.| **Priority:** ⬇️ Low **Estimated code review effort:** 3 (Moderate) | ~25 minutes **Change:** Refactor **Suggested reviewers:** `carlosmonastyrski`, `claude`, `andrewhuhh` Merge Risk: ⚪ Minimal · up to a0e48 The migration of the CA installation flow to v3 dialog components, plus the dialog layout option for app connection forms, shows no concrete merge-blocking issue. The sheet layout remains the default for existing callers. Security architecture risk: 🔵 Low · up to a0e48 The layout change preserves the inspected permission checks, project identity, and installation requests. No introduced security weakness was identified, but clipboard failure behavior and interruption during installation remain only partially validated. **Retained concerns** No architecture-level concerns identified. Security review details **Security Blast Radius** - _inferred_ — The inspected new layout caller does not expand connection-creation authority or request scope: it retains the current project identity, and connection creation forwards projectId through the existing mutation. This establishes frontend continuity, not independent proof of backend authorization. **Trust Boundaries and Controls** - _observed_ — The CA connection-creation affordance is unavailable unless the project permission permits creating app connections. Changing the layout does not replace that check or move completion and cancellation authority away from the surrounding caller. **Resilience and Maintainability Implications** - _observed_ — The existing connection-creation discard guard increments an attempt token on dismissal and ignores a completion from an invalidated attempt. The PR adds only the layout argument at this caller. The guard protects subsequent UI selection; it does not demonstrate cancellation or rollback of a server-side creation request. Pre-merge checks: 4 · 1 Failed checks: 1 warning | Check name | Status | Explanation | Resolution | | :---------------- | :----------------------------------------------------------------------------------------------: | :-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | :--------------------------------------------------------------------------------- | | Docstring Coverage | | Docstring coverage is 0.00% which is insufficient. The required threshold is 80.00%. Docstring coverage is scoped to functions touched by this diff. Analyzed 1 functions across 10 files. (1 skipped: 1… | Write docstrings for the functions missing them to satisfy the coverage threshold. | Passed checks: 4 passed | Check name | Status | Explanation | | :------------------------ | :--------------------------------------------------------------------------------------------: | :-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | Title check | | The title clearly and concisely describes the primary change: migrating CaInstallCertModal to the v3 Dialog. It uses the conventional commit format. | | Description check | | The description includes the required Context, Screenshots, Steps to verify the change, Type, and Checklist sections. It explains the migration, testing coverage, and compatibility with the existing s… | | Linked Issues check | | Check skipped because no linked issues were found for this pull request. | | Out of Scope Changes check | | Check skipped because no linked issues were found for this pull request. | Full details: Description check **Explanation** The description includes the required Context, Screenshots, Steps to verify the change, Type, and Checklist sections. It explains the migration, testing coverage, and compatibility with the existing sheet layout. Full details: Docstring Coverage **Explanation** Docstring coverage is 0.00% which is insufficient. The required threshold is 80.00%. Docstring coverage is scoped to functions touched by this diff. Analyzed 1 functions across 10 files. (1 skipped: 1 unsupported.) **Resolution** Write docstrings for the functions missing them to satisfy the coverage threshold. ✨ Finishing Touches: 💡 1 📝 Generate docstrings: 💡 🧪 Generate unit tests (beta) Comment `@the reviewer help` to get the list of available commands.
#8592
2026-10-20 15:41
cubic
review-body
—summary
**3 issues found** across 11 files Reply with feedback, questions, or to request a fix.View guided diff | Turn on auto-fix | Re-trigger the reviewer
#8592
2026-10-20 15:41
cubic
inline · P2
frontend/src/pages/cert-manager/CertificateAuthoritiesPage/components/CaInstallCertModal/InternalCaInstallForm.tsx:158defect-functional
These three labels are not programmatically connected to their sibling controls, leaving the parent, validity, and path-length controls without accessible names. Add a matching `htmlFor` and control `id` for each field.
#8592
2026-10-20 15:41
cubic
inline · P2
frontend/src/pages/cert-manager/CertificateAuthoritiesPage/components/CaInstallCertModal/ExternalCaInstallForm.tsx:88defect-functional
This renders the CSR section title as an unassociated `<label>` instead of a heading, so screen-reader heading navigation skips it. Use a heading element for this section title. ```suggestion <h3>Certificate Signing Request</h3> ```
#8592
2026-10-20 15:41
cubic
inline · P2
frontend/src/pages/organization/AppConnections/AppConnectionsPage/components/AppConnectionForm/AppConnectionFormFooter.tsx:37defect-functional
This fallback is the default for existing AppConnectionForm sheets, so it also changes their footer to put Cancel first and render it as ghost. Preserve the sheet's original submit-first, outline-Cancel layout; keep the new treatment in the dialog branch.
#8592
2026-10-20 15:47
greptile
issue-comment
—summary
Confidence Score: 5/5 **[Medium impact]** No merge-blocking issues were identified in this frontend dialog migration. Summary Migrates the CA certificate install flow from the v2 modal to the v3 dialog. - CA certificate installation now runs in a v3 dialog. - CA install forms use v3 fields and scrolling sections. - Connection setup opens as a dialog inside CA installation. No actionable issues were found. No merge-blocking issues were identified. Reviews (1) · Last reviewed commit: "improvement(cert-manager): migrate CaIns..." · Reviewed by the reviewer
#8568
2026-10-20 15:49
coderabbit
issue-comment
—summary
📝 Walkthrough Secret Scanning now uses organization-scoped URLs without project IDs. Backend project creation and project resolution establish the active project, while frontend cookies and navigation preserve project selection. Secret Scanning screens and shared access-management interfaces now use product-specific routing and labels. **Changes: Secret Scanning organization scope** |Layer / File(s)|Summary| |:---|:---| |**Project bootstrap and request resolution** `backend/src/ee/services/secret-scanning-v2/*`, `backend/src/ee/services/sub-org/sub-org-service.ts`, `backend/src/services/org/org-service.ts`, `backend/src/server/plugins/inject-secret-scanning-project-id*`, `backend/src/ser
… the rest (7275 characters)vices/project/*`, `backend/src/services/smtp/emails/SecretScanningScanFailedTemplate.tsx`, `backend/CLAUDE.md`, `CLAUDE.md`|Organization and sub-organization creation bootstrap Secret Scanning projects for creators. Request resolution honors explicit project IDs and valid organization-scoped cookies before falling back to the active project. Project billing counts and failure-email project details were updated.| |**Implicit project resolution and organization routes** `frontend/src/helpers/project.ts`, `frontend/src/helpers/secretScanningActiveProject.ts`, `frontend/src/hooks/*`, `frontend/src/pages/secret-scanning/layout.tsx`, `frontend/src/routes.ts`, `frontend/src/const/routes.ts`, `frontend/src/pages/redirects/*`, `frontend/src/pages/secret-scanning/*/route.tsx`, `frontend/src/pages/project/*/route-secret-scanning.tsx`|Secret Scanning routes no longer require `$projectId`. The layout resolves an implicit project from active-project and membership data. Index and legacy project-URL redirects preserve organization context, and legacy project URLs can pin the project cookie.| |**Project selection and navigation** `frontend/src/components/v3/platform/RootCommandMenu/RootCommandMenu.tsx`, `frontend/src/helpers/organizationSwitch.ts`, `frontend/src/layouts/ProjectLayout/components/TypeSelect/TypeSelect.tsx`, `frontend/src/layouts/SecretScanningLayout/components/SecretScanningInstanceBanner.tsx`, `frontend/src/pages/organization/ProjectsPage/*`, `frontend/src/pages/organization/SettingsPage/components/OrgSecretScanningTab/OrgSecretScanningTab.tsx`, `frontend/src/pages/organization/*/*ProjectRow.tsx`, `frontend/src/pages/public/ErrorPage/components/ProjectAccessError.tsx`|Project selection is stored in an organization-scoped cookie before navigation. Organization project views resolve the Secret Scanning project and handle access restrictions; project links and access-error resolution account for Secret Scanning.| |**Secret Scanning and access-management interface** `frontend/src/pages/secret-scanning/*`, `frontend/src/pages/project/AccessControlPage/*`, `frontend/src/pages/project/GroupDetailsByIDPage/*`, `frontend/src/pages/project/IdentityDetailsByIDPage/*`, `frontend/src/pages/project/MemberDetailsByIDPage/*`, `frontend/src/pages/project/RoleDetailsBySlugPage/*`, `frontend/src/pages/project/AppConnectionsPage/*`, `frontend/src/pages/project/AuditLogsPage/*`, `frontend/src/pages/organization/AppConnections/*`, `frontend/src/components/auth/signupProducts.ts`|Secret Scanning screens use organization-level navigation and revised product wording. Shared access-management labels now use product visibility and label helpers instead of Certificate Manager-only conditions.| **Priority:** ➖ Normal **Estimated code review effort:** 4 (Complex) | ~60 minutes **Change:** Bug fix ### Sequence Diagram(s) ```mermaid sequenceDiagram participant OrganizationService participant bootstrapSecretScanningProject participant Transaction OrganizationService->>bootstrapSecretScanningProject: bootstrap project for organization creator bootstrapSecretScanningProject->>Transaction: create project and administrator access ``` **Suggested reviewers:** `scott-ray-wilson`, `sheensantoscapadngan`, `akhilmhd` Merge Risk: 🔵 Low · up to e6432 Custom roles cannot be opened from a machine identity’s role card, and two messages still call Secret Scanning a project. These are bounded issues that can be fixed before merge or accepted for follow-up. Security architecture risk: 🔵 Low · up to e6432 The change affects project selection and administrator setup. Existing access checks and all-or-nothing provisioning limit the identified risk. No new exploitable path was established, although compatibility and operational coverage remain incomplete. **Retained concerns** No architecture-level concerns identified. Security review details **Security Blast Radius** - _observed_ — The new persistent authority consists of project-administrator membership in the newly created organization's Secret Scanning project. Backend cookie selection is constrained to a Secret Scanning project belonging to the authenticated organization. **Security Findings and Attack Paths** - _observed_ — A modified selection cookie can influence project selection but does not itself grant data-source creation authority. The inspected creation operation obtains project permission and checks the Create action before using connections, encrypting credentials, persisting the data source, or queuing a scan. These service controls are unchanged by the PR. **Trust Boundaries and Controls** - _observed_ — The sub-organization endpoint requires JWT or identity access-token authentication and passes server-derived permission context to the service. Parent-organization permission enforcement precedes the new bootstrap call; actor and organization binding remain unchanged from the comparison base. **Resilience and Maintainability Implications** - _observed_ — Existing database definitions assign cleanup ownership through cascading organization-to-project, organization/project-to-membership, and membership-to-role foreign keys. Organization deletion delegates the database removal within a transaction, providing an existing cleanup path for the newly provisioned records. Pre-merge checks: 5 | Check name | Status | Explanation | | :------------------------ | :--------------------------------------------------------------------------------------------: | :-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | Title check | | The title clearly summarizes the primary change: removing project terminology from Secret Scanning. It is concise and follows the conventional commit format. | | Description check | | The description explains the problem, affected areas, intended pattern, and migration behavior. The verification steps are empty, and checklist items remain unchecked, but the description is otherwise… | | Docstring Coverage | | No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check. Docstring coverage is scoped to functions touched by this diff. Analyzed 0 functions across 5… | | Linked Issues check | | Check skipped because no linked issues were found for this pull request. | | Out of Scope Changes check | | Check skipped because no linked issues were found for this pull request. | Full details: Description check **Explanation** The description explains the problem, affected areas, intended pattern, and migration behavior. The verification steps are empty, and checklist items remain unchecked, but the description is otherwise sufficiently complete. Full details: Docstring Coverage **Explanation** No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check. Docstring coverage is scoped to functions touched by this diff. Analyzed 0 functions across 50 files. (35 skipped: 2 unsupported, 33 over the file limit.) ✨ Finishing Touches 📝 Generate docstrings 🧪 Generate unit tests (beta) Comment `@the reviewer help` to get the list of available commands.
#8568
2026-10-20 16:03
coderabbit
inline · P2
frontend/src/pages/project/IdentityDetailsByIDPage/components/IdentityRoleDetailsSection/IdentityRoleDetailsSection.tsx:148defect-functional
**Do not treat Secret Scanning as a standalone product for role behavior.** `isStandaloneProduct` controls more than the card title. It also does these things: - Role rows become non-clickable, with no `role`, no `tabIndex`, and no `onClick`. - The sheet description says "pre-defined roles" only. Secret Scanning supports custom project roles. `ProjectRoleList` shows "Add Role" for every type except Certificate Manager, and this PR adds `RoleDetailsBySlugPage/route-secret-scanning.tsx`. After this change, a machine identity's custom Secret Scanning role cannot be opened from this card. `MemberRoleDetailsSection` still allows that navigation for users, so the two pages now behave differently
… the rest (611 characters). Use `hidesProjects` for the title only. Keep navigation and the sheet text tied to `isCertManager || isPam`. Proposed fix ```diff - const isSecretScanning = currentProject?.type === ProjectType.SecretScanning; - const isStandaloneProduct = isCertManager || isPam || isSecretScanning; + const isStandaloneProduct = isCertManager || isPam; + const projectsHidden = hidesProjects(currentProject?.type); ``` ```diff - <CardTitle>{isStandaloneProduct ? "Roles" : "Project Roles"}</CardTitle> + <CardTitle>{isStandaloneProduct || projectsHidden ? "Roles" : "Project Roles"}</CardTitle> ```
#8568
2026-10-20 16:03
coderabbit
review-body
—summary
**Actionable comments posted: 1** > [!CAUTION] > Some comments are outside the diff and can’t be posted inline due to GitHub limitations. > > **⚠️ Outside diff range comments (2)** > > > 🟡 Minor · Use getProjectLabel in the remove-user success toast. · MemberDetailsByIDPage.tsx:107-112 > > `frontend/src/pages/project/MemberDetailsByIDPage/MemberDetailsByIDPage.tsx:107-112` > **🎯 Functional Correctness** | **🟡 Minor** | **⚡ Quick win** > > **Use `getProjectLabel` in the remove-user success toast.** > > This toast still checks only `ProjectType.CertificateManager`. When a user is removed on Secret Scanning, the toast says "Successfully removed user from project". The other strings in this
… the rest (1730 characters)file now use `projectLabel`. Use the same helper here. You can compute the label once above `handleRemoveUser`. > > > Proposed fix > > ```diff > - text: `Successfully removed user from ${ > - currentProject?.type === ProjectType.CertificateManager ? "Certificate Manager" : "project" > - }`, > + text: `Successfully removed user from ${ > + hidesProjects(currentProject?.type) ? getProjectLabel(currentProject?.type) : "project" > + }`, > ``` > > > > > > > > > 🟡 Minor · Use the product label in the delete-identity error and… · IdentityDetailsByIDPage.tsx:210-215 > > `frontend/src/pages/project/IdentityDetailsByIDPage/IdentityDetailsByIDPage.tsx:210-215` > **🎯 Functional Correctness** | **🟡 Minor** | **⚡ Quick win** > > **Use the product label in the delete-identity error and dialog.** > > The rest of this page now uses `accessControlLabel` and `namesProduct`. The delete-identity error toast and the delete dialog description still check only `isCertManager`. On Secret Scanning, users see "Failed to delete project machine identity" and "permanently deletes the project machine identity". This goes against the PR goal of removing "project" wording from Secret Scanning. Use `getProjectLabel(currentProject?.type)` when `projectsHidden` is true. > > > Proposed fix > > ```diff > - text: isCertManager > - ? "Failed to delete Certificate Manager machine identity" > - : "Failed to delete project machine identity" > + text: `Failed to delete ${projectsHidden ? getProjectLabel(currentProject?.type) : "project"} machine identity` > ... > - isCertManager ? "Certificate Manager" : "project" > + projectsHidden ? getProjectLabel(currentProject?.type) : "project" > ``` > > > > > > > > Also applies to: 457-459 > > > > > >
#8568
2026-10-20 16:04
greptile
issue-comment
—summary
Confidence Score: 4/5 **[Medium impact]** The changes affect Secret Scanning entry\, access\, and project setup across organizations\. Not safe to merge until navigation can no longer silently change a tab's selected project and redirect subsequent edits. Findings 1. **Tabs switch projects unexpectedly** Summary Removes project wording from Secret Scanning and moves its pages to organization-scoped URLs. - Secret Scanning opens at an organization URL and keeps its selected project in scope. - Secret Scanning access, connection, and scan screens use product-focused wording. - New organizations get a Secret Scanning project with creator admin access. Do not merge until project selection
… the rest (201 characters) remains stable within each tab; the current behavior can apply edits to the wrong project. Reviews (1) · Last reviewed commit: "Merge branch 'main' into matheus/remove-..." · Reviewed by the reviewer
#8568
2026-10-20 16:04
greptile
inline · P1
frontend/src/pages/secret-scanning/layout.tsx:39defect-functional
**Tabs switch projects unexpectedly** `beforeLoad` reads the shared cookie on each navigation. In an organization with several legacy projects, opening project A in one tab and then project B from organization settings in another tab overwrites that cookie. Navigating to Findings or Settings in the first tab then switches its project to B without an explicit project-selection action. Saving configuration in that tab sends the update to B instead of A. Explicit `projectId` values do not prevent this: navigation has already replaced the project context and loaded B’s configuration. Keep the viewed project stable per tab, changing it only when the user explicitly opens another project. This n
… the rest (1735 characters)eeds to be fixed before merging because ordinary navigation can redirect edits to the wrong project. Artifacts **Evidence from the check** - Contains the complete executed harness, narrow fixtures, imported product code, synchronization, mutation assertions, and capture commands; both comparison runs now complete successfully. **Command output from the check** - Records the command, working directory, exit code 0, settled route contexts, and request traces for both runs; head changes tab A’s write target from A to B. **Recording of the check** - Records tab A opening A, tab B selecting B, and tab A navigating through Findings and Settings before saving; the historical route preserves project A. **Tab A after saving with the historical project-ID route** - Captures the rendered Settings form after tab A completes the comparison save; the selected project remains A. **Tab B after selecting project B in the historical comparison** - Captures tab B after the actual organization-settings selector opens B; the second-tab selection was exercised before tab A navigated. **Recording of the check** - Records tab A opening A, tab B selecting B, and tab A settling on B through Findings and Settings before saving; the changed route redirects the subsequent edit to B. **Tab A after saving project B configuration at head** - Captures the rendered Settings form after tab A’s originally project-A session saves following the cross-tab selection; the settled selected project is B. **Tab B after selecting project B at head** - Captures tab B after the real organization-settings action opens B and writes the shared cookie; this selection precedes tab A’s context replacement. Ran code and verified through TREX
#8568
2026-10-20 16:08
cubic
review-body
—summary
**11 issues found** across 88 files Reply with feedback, questions, or to request a fix.View guided diff | Turn on auto-fix | Re-trigger the reviewer
#8568
2026-10-20 16:08
cubic
inline · P1
frontend/src/pages/secret-scanning/layout.tsx:34defect-functional
The layout re-reads the organization-wide cookie on navigation, so another tab selecting project B can switch this tab from A and send configuration edits to B. Keep this tab’s resolved project stable, changing it only after an explicit project selection.
#8568
2026-10-20 16:08
cubic
inline · P2
frontend/src/pages/secret-scanning/SettingsPage/SettingsPage.tsx:38defect-functional
This bypasses i18n for the browser title, so users of supported non-English locales now see “Secret Scanning Settings” in English. Add a localized Secret Scanning settings title key and pass its `t(...)` result here.
#8568
2026-10-20 16:08
cubic
inline · P2
CLAUDE.md:116docs
This overstates when new organizations are bootstrapped: orgs created from the admin console have no creator, so `org-service.ts` skips this call and they resolve the project lazily. Qualify the creation claim and include admin-console orgs in the lazy-resolution case. ```suggestion Secret Scanning has one active project per organization (the newest). Like Certificate Manager, its URLs are `/organizations/$orgId/secret-scanning/…` with no `$projectId`; legacy orgs can still hold several projects, so the frontend pins the one being viewed (see `frontend/CLAUDE.md`). New orgs with a creator and sub-orgs get it at creation (`bootstrapSecretScanningProject`); older orgs and orgs created from th
… the rest (229 characters)e admin console get it lazily from `GET /api/v2/secret-scanning/project`. `projectId` is optional on its API (`inject-secret-scanning-project-id`). Do not add a project picker or a way to create more Secret Scanning projects. ```
#8568
2026-10-20 16:08
cubic
inline · P2
frontend/src/pages/project/AccessControlPage/components/ProjectRoleListTab/components/ProjectRoleList/ProjectRoleList.tsx:89defect-functional
Secret Scanning still shows project-specific copy in this card and after role deletion. Update these `isCertManager` copy checks to use the projectless-product condition so the Access Control page consistently removes “project” language.
#8568
2026-10-20 16:08
cubic
inline · P2
frontend/src/pages/organization/ProjectsPage/components/ProjectCategoryOverview.tsx:431defect-functional
`ProjectCategoryOverview` ignores the per-org pinned project, so members of a legacy Secret Scanning project can have the tile blocked or be sent to the active project. Resolve the cookie against `projects` before computing access and entering.
#8568
2026-10-20 16:08
cubic
inline · P2
frontend/src/helpers/organizationSwitch.ts:65defect-functional
This detour replaces a legacy organization’s saved Secret Scanning selection with its active project before opening the page. Route directly to `/secret-scanning/data-sources` so the layout’s cookie-aware resolver preserves the selected project. ```suggestion section = "secret-scanning/data-sources"; ```
#8568
2026-10-20 16:08
cubic
inline · P2
backend/src/services/org/org-service.ts:811defect-functional
The instance-bootstrap identity is created as an org admin only after this call, so it never receives Secret Scanning project membership. Project permission checks require that membership, and the lazy resolver will not backfill it once this project exists; add project-admin membership when the identity is created so instance-admin identity requests can access Secret Scanning.
#8568
2026-10-20 16:08
cubic
inline · P2
frontend/src/helpers/secretScanningActiveProject.ts:35defect-functional
This accepts any project in the user's global project list, not just a Secret Scanning project in `orgId`. A legacy URL can pin another organization's or product's UUID, causing the Secret Scanning layout to resolve to the wrong project; validate against org- and product-scoped membership before returning it.
#8568
2026-10-20 16:08
cubic
inline · P2
frontend/src/components/v3/platform/RootCommandMenu/RootCommandMenu.tsx:869defect-functional
These links omit both the URL project ID and a cookie update. In a legacy multi-project org, a cookie change from another tab can make a later layout load resolve a different project than this menu's `project`; pin `project.id` before each navigation.
#8568
2026-10-20 16:08
cubic
inline · P2
frontend/src/pages/secret-scanning/layout.tsx:21defect-functional
`ensureQueryData` returns a cached active project ID even after it becomes stale, so revisiting after the org’s active Secret Scanning project changes can keep loading the previous project. Fetch this org-scoped ID on route load instead of accepting stale cached data. ```suggestion context.queryClient.fetchQuery<string>({ queryKey: secretScanningV2Keys.activeProjectId(params.orgId), queryFn: fetchSecretScanningProjectId, staleTime: 0 }), ```
#8568
2026-10-20 16:08
cubic
inline · P3
frontend/src/pages/project/IdentityDetailsByIDPage/components/IdentityRoleDetailsSection/IdentityRoleDetailsSection.tsx:148docs
This changes the editor description to say only pre-defined roles are available, but the editor also offers custom roles for Secret Scanning. Keep the custom-role wording for this product.
#8614
2026-10-20 16:10
coderabbit
issue-comment
—summary
📝 Walkthrough The PR migrates secret-manager integration authorization and configuration pages from v2 to v3 UI components. Forms use structured cards, labeled fields, inline validation, and updated pending-button props. Several authorization and integration-creation flows now clear loading state in `finally` blocks. **Changes: Integration authorization and configuration** |Layer / File(s)|Summary| |:---|:---| |**Authorization forms** `frontend/src/pages/secret-manager/integrations/*AuthorizePage/*AuthorizePage.tsx`|Authorization pages use v3 card and field components, documentation badges, accessible validation messages, and pending-aware buttons. Several flows now clear loading state i
… the rest (4378 characters)n `finally` blocks.| |**Configuration forms** `frontend/src/pages/secret-manager/integrations/*ConfigurePage/*ConfigurePage.tsx`|Configuration pages use v3 card, field, tab, select, toggle, and combobox components. Existing form values and conditional behavior remain wired through the new components. Several flows now clear loading state in `finally` blocks.| **Priority:** ➖ Normal **Estimated code review effort:** 3 (Moderate) | ~25 minutes **Change:** Refactor **Suggested reviewers:** `andrewhuhh` Merge Risk: 🔵 Low · up to e7218 The integration pages move to the newer UI components and keep their existing submission behavior. Loading spinners now also clear after failed requests. One small issue remains on the Azure DevOps connect page: a missing API token error also appears, and is announced, under the organization-name field. The change is safe to merge once that wiring is fixed or accepted. Security architecture risk: 🔵 Low · up to e7218 The inspected flows preserve project scope, credential selection and submission payloads while improving recovery after failed submissions. Remaining uncertainty concerns affected flows that were not fully traced. **Retained concerns** No architecture-level concerns identified. Security review details **Security Blast Radius** - _inferred_ — The security-sensitive outcomes remain creation of project integration credentials and configuration of secret synchronization to selected provider targets. The inspected migration does not introduce another service, credential authority or target-selection mechanism; this conclusion is limited to the traced flows. **Trust Boundaries and Controls** - _observed_ — The inspected Checkly route retains authentication and organization/project context middleware. Its access-token request still carries the current project identifier, while the backend authenticates the request and enforces project integration-creation permission. Client pending state is not the authorization boundary. **Resilience and Maintainability Implications** - _observed_ — V3 pending buttons disable ordinary repeated activation, and finally cleanup releases the local lock after failure. This does not establish server-side idempotency or roll back persisted credentials. The inspected persistence operation still creates records, and successful authorization still invokes navigation without awaiting it; both behaviors predate this PR. Pre-merge checks: 4 · 1 Failed checks: 1 warning | Check name | Status | Explanation | Resolution | | :---------------- | :----------------------------------------------------------------------------------------------: | :----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | :--------------------------------------------------------------------------------- | | Docstring Coverage | | Docstring coverage is 0.00% which is insufficient. The required threshold is 80.00%. Docstring coverage is scoped to functions touched by this diff. Analyzed 1 functions across 16 files. | Write docstrings for the functions missing them to satisfy the coverage threshold. | Passed checks: 4 passed | Check name | Status | Explanation | | :------------------------ | :--------------------------------------------------------------------------------------------: | :------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | Title check | | The title clearly identifies the migration of native secret-manager integration creation pages to V3 and follows the conventional commit format. | | Description check | | The description includes the required Context, Screenshots, Steps to verify the change, Type, and Checklist sections. It provides detailed scope, verification results, limitations, and testing status. | | Linked Issues check | | Check skipped because no linked issues were found for this pull request. | | Out of Scope Changes check | | Check skipped because no linked issues were found for this pull request. | ✨ Finishing Touches: 💡 1 📝 Generate docstrings: 💡 🧪 Generate unit tests (beta) Comment `@the reviewer help` to get the list of available commands.
#8614
2026-10-20 16:15
coderabbit
inline · P2
frontend/src/pages/secret-manager/integrations/AzureDevopsAuthorizePage/AzureDevopsAuthorizePage.tsx:160defect-functional
**The Organization Name field shows the API token error as its own error.** The Organization Name `Field` reads its state from `apiKeyErrorText`. This affects `data-invalid`, `isError`, `aria-describedby` and the `FieldError` content. If the API token is blank, the page shows "API Key cannot be blank" twice: once under the token field and once under the organization name field. Both `FieldError` elements render with `role="alert"`, so screen readers announce the message twice. The organization-name input's accessible description also says the API key is blank, which is wrong for that input. `handleButtonClick` does not validate `devopsOrgName`. Remove the error wiring from the organization
… the rest (2120 characters)-name field. Add a separate `devopsOrgNameErrorText` state only if this field needs its own validation. 🐛 Proposed fix ```diff - <Field data-invalid={apiKeyErrorText !== ""}> + <Field> <FieldLabel htmlFor="azuredevops-azure-devops-organization-name"> @@ <Input placeholder="" value={devopsOrgName} onChange={(e) => setDevopsOrgName(e.target.value)} id="azuredevops-azure-devops-organization-name" - isError={apiKeyErrorText !== ""} - aria-describedby={ - apiKeyErrorText !== "" - ? "azuredevops-azure-devops-organization-name-error" - : undefined - } /> - {apiKeyErrorText !== "" && ( - <FieldError id="azuredevops-azure-devops-organization-name-error"> - {apiKeyErrorText} - </FieldError> - )} </Field> ``` 📝 Committable suggestion > ‼️ **IMPORTANT** > Carefully review the code before committing. Ensure that it accurately replaces the highlighted code, contains no missing lines, and has no issues with indentation. Thoroughly test & benchmark the code to ensure it meets the requirements. ```suggestion <Field> <FieldLabel htmlFor="azuredevops-azure-devops-organization-name"> Azure DevOps Organization Name <Tooltip> <TooltipTrigger asChild> <button type="button" aria-label="More information"> ? </button> </TooltipTrigger> <TooltipContent> This is the slug of the organization. An example would be &apos;my-acme-org&apos; </TooltipContent> </Tooltip> </FieldLabel> <Input placeholder="" value={devopsOrgName} onChange={(e) => setDevopsOrgName(e.target.value)} id="azuredevops-azure-devops-organization-name" /> </Field> ```
#8614
2026-10-20 16:15
coderabbit
review-body
—summary
**Actionable comments posted: 1**
#8614
2026-10-20 16:20
cubic
inline · P2
frontend/src/pages/secret-manager/integrations/AzureDevopsAuthorizePage/AzureDevopsAuthorizePage.tsx:134defect-functional
This tooltip trigger is a `<button>` nested inside the organization input’s `<label>`, which is invalid label content and can pollute the field’s accessible name. Render the label and tooltip trigger as siblings instead.
#8614
2026-10-20 16:20
cubic
inline · P2
frontend/src/pages/secret-manager/integrations/AzureDevopsAuthorizePage/AzureDevopsAuthorizePage.tsx:156defect-functional
Remove the API-key error wiring from the organization-name field; a blank token currently renders the same alert under both inputs and gives the organization input an incorrect accessible description.
#8614
2026-10-20 16:20
cubic
review-body
—summary
**2 issues found** across 16 files Reply with feedback, questions, or to request a fix.View guided diff | Turn on auto-fix | Re-trigger the reviewer
#8614
2026-10-20 16:23
greptile
issue-comment
—summary
Confidence Score: 5/5 **[Medium impact]** No merge-blocking issue was identified; focused browser checks support preserved form behavior, without establishing authenticated backend or provider acceptance. Summary Migrates 16 native-integration creation pages to V3 components. - Authorization pages show provider fields in V3 cards with connect actions in the footer. - Setup pages use V3 fields and controls while keeping provider choices in their existing flow. No actionable issues found. Focused browser checks of the actual AWS Parameter Store and CircleCI components preserved required validation, focus, form submission, dependent selections, and base/head mutation payloads. These checks
… the rest (590 characters)used synthetic application boundaries and do not establish authenticated backend or provider acceptance. The prior source-based checks and passing whitespace check remain applicable. PierreOucif explicitly identifies the inherited SecretPathInput ref and blur omissions as outside this migration. PierreOucif also intentionally preserves direct routes, copy, retirement behavior, and the current shared Card API. No merge-blocking issue was identified within the reviewed scope. Reviews (1) · Last reviewed commit: "improvement(secret-manager): migrate nat..." · Reviewed by the reviewer
#8617
2026-10-20 16:25
coderabbit
issue-comment
—summary
No actionable comments were generated in the recent review. 🎉 📝 Walkthrough Authorization and configuration pages for nine secret manager integrations now use v3 UI components. The forms add structured labels and validation messages, and several submission handlers reset loading state in `finally` blocks. Existing integration and selection flows are retained. **Changes: Secret manager integration pages** |Layer / File(s)|Summary| |:---|:---| |**Authorization forms** `frontend/src/pages/secret-manager/integrations/*AuthorizePage/*.tsx`|The authorization pages use v3 cards, fields, and pending-state buttons. Fields display labels and inline validation errors with accessibility associatio
… the rest (3369 characters)ns. Several authorization handlers now clear loading state in `finally` blocks.| |**Standard configuration forms** `frontend/src/pages/secret-manager/integrations/CloudflarePagesConfigurePage/CloudflarePagesConfigurePage.tsx`, `.../CloudflareWorkersConfigurePage/CloudflareWorkersConfigurePage.tsx`, `.../CodefreshConfigurePage/CodefreshConfigurePage.tsx`, `.../DatabricksConfigurePage/DatabricksConfigurePage.tsx`, `.../DigitalOceanAppPlatformConfigurePage/DigitalOceanAppPlatformConfigurePage.tsx`, `.../FlyioConfigurePage/FlyioConfigurePage.tsx`|These forms use v3 card sections, labeled fields, updated select controls, and pending-state actions. Existing selection behavior and conditions that disable actions when no services or apps are available remain in place. Several handlers now clear loading state in `finally` blocks.| |**GitHub and GitLab configuration forms** `frontend/src/pages/secret-manager/integrations/GithubConfigurePage/GithubConfigurePage.tsx`, `frontend/src/pages/secret-manager/integrations/GitlabConfigurePage/GitlabConfigurePage.tsx`|The forms use v3 fields, tabs, selectors, toggles, and dialogs. Submission handlers clear loading state in `finally` blocks. The GitLab page no longer destructures `handlePopUpClose` from the popup hook.| **Priority:** ➖ Normal **Estimated code review effort:** 4 (Complex) | ~45 minutes **Change:** Refactor **Suggested reviewers:** `andrewhuhh` Merge Risk: ⚪ Minimal · up to 39b98 The integration authorization and configuration pages move to the newer UI components. Their submit actions, OAuth flows, and selection behavior remain wired to the existing handlers. No concrete issue blocks merging. Provider submit and OAuth paths were not exercised end to end. Pre-merge checks: 5 | Check name | Status | Explanation | | :------------------------ | :--------------------------------------------------------------------------------------------: | :-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | Title check | | The title clearly and concisely describes the migration of native secret-manager integration creation pages to V3 and follows the conventional commit format. | | Description check | | The description is complete and relevant. It documents the migration scope, preserved behavior, UI changes, screenshots, verification steps, change type, checklist status, and testing limitations. | | Docstring Coverage | | No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check. Docstring coverage is scoped to functions touched by this diff. Analyzed 0 functions across 1… | | Linked Issues check | | Check skipped because no linked issues were found for this pull request. | | Out of Scope Changes check | | Check skipped because no linked issues were found for this pull request. | Full details: Docstring Coverage **Explanation** No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check. Docstring coverage is scoped to functions touched by this diff. Analyzed 0 functions across 17 files. ✨ Finishing Touches 📝 Generate docstrings 🧪 Generate unit tests (beta) Comment `@the reviewer help` to get the list of available commands.
#8617
2026-10-20 16:35
cubic
inline · P2
frontend/src/pages/secret-manager/integrations/GcpSecretManagerAuthorizePage/GcpSecretManagerAuthorizePage.tsx:160defect-functional
The field loses its previous `h-48` and now uses V3's `min-h-16`, leaving the service-account JSON editor only 4rem tall. Restore `h-48` so users can review pasted credentials without constant scrolling. ```suggestion <TextArea className="h-48" ```
#8617
2026-10-20 16:35
cubic
review-body
—summary
**1 issue found** across 17 files Reply with feedback, questions, or to request a fix.View guided diff | Turn on auto-fix | Re-trigger the reviewer
#8617
2026-10-20 16:42
greptile
issue-comment
—summary
Confidence Score: 5/5 **[Medium impact]** No blocking issue was identified; the changes are safe to merge within the reviewed scope. Summary Migrates 17 native-integration creation pages from V2 to V3 components, adding composed card headers, content, footers, fields, menus, tabs, and toggles. Loading-state cleanup moves into `finally`, and GitLab’s overwrite prompt is protected while pending. No actionable regression was found. Browser checks of the actual GitHub page source with synthetic project context and API responses confirmed repository selection and menu dismissal, value preservation across tabs, footer submission, pending duplicate-click prevention, and successful retry after r
… the rest (871 characters)ejection. No real provider writes, external OAuth, or authenticated application routing were exercised. The other 16 pages were not rendered during these checks; the established source checks for GCP’s separate OAuth action and GitLab’s pending confirmation remain preserved. Acknowledged by `PierreOucif`: `SecretPathInput` ref/blur omissions and GitLab’s `initialSyncBehavior` ref/blur omissions are inherited limitations. Shared footer-band styling is explicitly deferred. No blocking issue was identified within the reviewed scope. - Provider credential pages use the V3 card layout. - Provider configuration pages use V3 cards and footers. - GitHub setup keeps its tabs and choices in a V3 card. - GitLab setup keeps its tabs and overwrite prompt in V3. Reviews (1) · Last reviewed commit: "improvement(secret-manager): migrate nat..." · Reviewed by the reviewer
#8616
2026-10-20 17:16
coderabbit
issue-comment
—summary
📝 Walkthrough Secret Manager integration authorization and configuration pages now use v3 UI components. Several pages derive submit-button loading feedback from mutation pending state. The GCP configuration page also replaces its overwrite-warning modal with a v3 dialog. **Changes: Authorization forms** |Layer / File(s)|Summary| |:---|:---| |**Authorization forms** `frontend/src/pages/secret-manager/integrations/*AuthorizePage/*.tsx`|Provider authorization forms use v3 cards, fields, validation feedback, and pending-aware buttons. Several pages now use mutation pending state instead of local loading state.| **Configuration forms** |Layer / File(s)|Summary| |:---|:---| |**Configuration
… the rest (4848 characters) forms and pending state** `frontend/src/pages/secret-manager/integrations/HashicorpVaultConfigurePage/*.tsx`, `HasuraCloudConfigurePage/*.tsx`, `HerokuConfigurePage/*.tsx`, `LaravelForgeConfigurePage/*.tsx`, `NetlifyConfigurePage/*.tsx`, `NorthflankConfigurePage/*.tsx`, `RailwayConfigurePage/*.tsx`|These pages use v3 form and card components. Existing configuration fields and selection behavior remain; several pages use mutation pending state for submit feedback.| |**Multi-select configuration forms** `frontend/src/pages/secret-manager/integrations/OctopusDeployConfigurePage/*.tsx`, `QoveryConfigurePage/*.tsx`|These pages use v3 fields, cards, and selection components while retaining their provider configuration controls and pending-aware submit buttons.| |**GCP configuration tabs and overwrite dialog** `frontend/src/pages/secret-manager/integrations/GcpSecretManagerConfigurePage/GcpSecretManagerConfigurePage.tsx`|The connection and options tabs use v3 components. The overwrite-warning prompt is now a v3 dialog with Cancel and Continue actions.| **Priority:** ⬇️ Low **Estimated code review effort:** 4 (Complex) | ~45 minutes **Change:** Refactor **Suggested reviewers:** `andrewhuhh` Merge Risk: 🔵 Low · up to ff5f6 The migration keeps the existing submit behavior. Remaining issues are small wording and accessibility fixes on the GCP and Qovery pages, and they can be fixed before or shortly after merge. Security architecture risk: 🔵 Low · up to ff5f6 The inspected changes preserve credential submission, secret-source selection, provider targets, and success navigation. Pending-aware actions improve request lifecycle handling. No introduced security concern was established, but server-side authorization and duplicate-request recovery were not independently verified. **Retained concerns** No architecture-level concerns identified. Security review details **Security Blast Radius** - _inferred_ — The inspected frontend delta continues to act through existing project authorization records and selected provider resources. No additional credential authority or broader target scope was identified in the GCP and Octopus comparisons; server-enforced isolation was not independently established. **Trust Boundaries and Controls** - _observed_ — User-entered credentials are submitted with the current project identifier to the existing integration-auth endpoint. Configuration sends an integration authorization identifier and selected synchronization scope to the existing integration endpoint. These client-side selections and pending controls do not establish server-side authorization. **Resilience and Maintainability Implications** - _observed_ — GCP’s absence of an explicit form-level in-flight submission guard predates this PR. Pending buttons reduce repeat interaction but do not prove duplicate-request safety or reconciliation after an interrupted request. No introduced worsening was established. Pre-merge checks: 5 | Check name | Status | Explanation | | :------------------------ | :--------------------------------------------------------------------------------------------: | :-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | Title check | | The title clearly and concisely describes the main change: migrating native provider creation pages to V3. It uses the conventional commit format. | | Description check | | The description is complete and follows the repository template. It includes context, screenshots, verification steps, type classification, checklist status, scope, preserved behavior, and known verif… | | Docstring Coverage | | No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check. Docstring coverage is scoped to functions touched by this diff. Analyzed 0 functions across 1… | | Linked Issues check | | Check skipped because no linked issues were found for this pull request. | | Out of Scope Changes check | | Check skipped because no linked issues were found for this pull request. | Full details: Description check **Explanation** The description is complete and follows the repository template. It includes context, screenshots, verification steps, type classification, checklist status, scope, preserved behavior, and known verification limits. Full details: Docstring Coverage **Explanation** No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check. Docstring coverage is scoped to functions touched by this diff. Analyzed 0 functions across 17 files. ✨ Finishing Touches 📝 Generate docstrings 🧪 Generate unit tests (beta) Comment `@the reviewer help` to get the list of available commands.
#8616
2026-10-20 17:20
coderabbit
inline · P2
frontend/src/pages/secret-manager/integrations/GcpSecretManagerConfigurePage/GcpSecretManagerConfigurePage.tsx:454defect-functional
**Rename the `labelValue` field label to "Label Value".** The `labelValue` controller displays the text "Label Name". The `labelName` field on Line 430 uses the same text. When `shouldLabel` is on, the user sees two fields with the same label. The user cannot tell which input sets the GCP label value. Screen readers announce the same name for both inputs. 🐛 Proposed fix ```diff --- "a/frontend/src/pages/secret-manager/integrations/GcpSecretManagerConfigurePage/GcpSecretManagerConfigurePage.tsx" +++ "b/frontend/src/pages/secret-manager/integrations/GcpSecretManagerConfigurePage/GcpSecretManagerConfigurePage.tsx" @@ -449,9 +449,9 @@ name="labelValue"
… the rest (1006 characters) render={({ field, fieldState: { error } }) => ( <Field data-invalid={Boolean(error)}> <FieldLabel htmlFor="GcpSecretManagerConfigurePage-field-7"> - Label Name + Label Value </FieldLabel> <Input id="GcpSecretManagerConfigurePage-field-7" aria-describedby="GcpSecretManagerConfigurePage-field-7-feedback" ``` 📝 Committable suggestion > ‼️ **IMPORTANT** > Carefully review the code before committing. Ensure that it accurately replaces the highlighted code, contains no missing lines, and has no issues with indentation. Thoroughly test & benchmark the code to ensure it meets the requirements. ```suggestion <FieldLabel htmlFor="GcpSecretManagerConfigurePage-field-7"> Label Value </FieldLabel> ```
#8616
2026-10-20 17:20
coderabbit
inline · P2
frontend/src/pages/secret-manager/integrations/GcpSecretManagerConfigurePage/GcpSecretManagerConfigurePage.tsx:514defect-functional
**Add a `DialogDescription` to the overwrite-warning dialog.** `DialogContent` wraps Radix `DialogPrimitive.Content`. The dialog has a `DialogTitle` but no `DialogDescription`, so Radix does not set `aria-describedby` to the warning text. Assistive technology announces "Heads Up" but not the overwrite warning. Radix also logs a missing-description console warning in development. Use the v3 `DialogDescription` export, or set `aria-describedby` on `DialogContent` to the id of the warning paragraph. ♿ Proposed fix ```diff --- "a/frontend/src/pages/secret-manager/integrations/GcpSecretManagerConfigurePage/GcpSecretManagerConfigurePage.tsx" +++ "b/frontend/src/pages/secret-manager/integrations
… the rest (1249 characters)/GcpSecretManagerConfigurePage/GcpSecretManagerConfigurePage.tsx" @@ -502,16 +502,18 @@ onOpenChange={(isOpen) => handlePopUpToggle("confirmIntegration", isOpen)} > <DialogContent isPending={isLoading}> <DialogHeader> <DialogTitle>Heads Up</DialogTitle> + <DialogDescription> + You&apos;re about to overwrite any existing secrets in GCP Secret Manager. + </DialogDescription> </DialogHeader> <DialogBody> - <p>You&apos;re about to overwrite any existing secrets in GCP Secret Manager.</p> - <p className="mt-4"> + <p> To avoid this behavior, you may consider adding a secret prefix/suffix or enabling labeling in the options tab. </p> </DialogBody> <DialogFooter> <DialogClose asChild> <Button onClick={() => handlePopUpClose("confirmIntegration")} variant="ghost"> ``` Also add `DialogDescription` to the `@app/components/v3` import. Based on learnings: for Dialog overlays, "verify that ... proper ARIA roles/attributes (role="dialog", aria-modal, aria-labelledby, aria-describedby) are set." _Source: Learnings_
#8616
2026-10-20 17:20
coderabbit
inline · P2
frontend/src/pages/secret-manager/integrations/QoveryConfigurePage/QoveryConfigurePage.tsx:210defect-functional
**Replace "Checkly" with "Qovery" in the card description.** The Qovery configuration page tells users that their secrets sync to "Checkly environment variables". This is the wrong provider name, and users see it on every visit. This PR rewrites the line, so this is a good time to correct it. ✏️ Proposed fix ```diff - Choose which environment in Infisical you want to sync to Checkly environment variables. + Choose which environment in Infisical you want to sync to Qovery environment variables. ``` 📝 Committable suggestion > ‼️ **IMPORTANT** > Carefully review the code before committing. Ensure that it accurately replaces the highlighted code, contains no missing l
… the rest (291 characters)ines, and has no issues with indentation. Thoroughly test & benchmark the code to ensure it meets the requirements. ```suggestion <CardDescription> Choose which environment in Infisical you want to sync to Qovery environment variables. </CardDescription> ```
#8616
2026-10-20 17:20
coderabbit
inline · P2
frontend/src/pages/secret-manager/integrations/QoveryConfigurePage/QoveryConfigurePage.tsx:326docs
**Fix the typo "organizaztions" in the empty-state option.** Users see this text when they have no Qovery organizations. ✏️ Proposed fix ```diff --- "a/frontend/src/pages/secret-manager/integrations/QoveryConfigurePage/QoveryConfigurePage.tsx" +++ "b/frontend/src/pages/secret-manager/integrations/QoveryConfigurePage/QoveryConfigurePage.tsx" @@ -323,7 +323,7 @@ )) ) : ( <SelectItem value="none" key="target-app-none"> - No organizaztions found + No organizations found </SelectItem> )} <
… the rest (377 characters)/SelectContent> ``` 📝 Committable suggestion > ‼️ **IMPORTANT** > Carefully review the code before committing. Ensure that it accurately replaces the highlighted code, contains no missing lines, and has no issues with indentation. Thoroughly test & benchmark the code to ensure it meets the requirements. ```suggestion No organizations found ```
#8616
2026-10-20 17:20
coderabbit
review-body
—summary
**Actionable comments posted: 4**
#8616
2026-10-20 17:27
cubic
inline · P2
frontend/src/pages/secret-manager/integrations/NorthflankConfigurePage/NorthflankConfigurePage.tsx:126defect-functional
This `aria-describedby` points to an element that this page never renders; the other three controls have the same dangling reference. Remove these attributes or render matching feedback elements.
#8616
2026-10-20 17:27
cubic
inline · P2
frontend/src/pages/secret-manager/integrations/HashicorpVaultAuthorizePage/HashicorpVaultAuthorizePage.tsx:109defect-functional
Each input references a feedback ID even when its `FieldError` is not mounted, leaving dangling `aria-describedby` references before validation errors occur. Set each prop only when that field has an error.
#8616
2026-10-20 17:27
cubic
review-body
—summary
**9 issues found** across 17 files Reply with feedback, questions, or to request a fix.View guided diff | Turn on auto-fix | Re-trigger the reviewer
#8616
2026-10-20 17:27
cubic
inline · P2
frontend/src/pages/secret-manager/integrations/LaravelForgeConfigurePage/LaravelForgeConfigurePage.tsx:106defect-functional
These `aria-describedby` targets are never rendered, so the controls expose broken ARIA references. Remove the three attributes or render matching feedback elements.
#8616
2026-10-20 17:27
cubic
inline · P2
frontend/src/pages/secret-manager/integrations/RailwayConfigurePage/RailwayConfigurePage.tsx:151defect-functional
These `aria-describedby` references point to elements this page never renders, so assistive technology receives no described-by text and accessibility checks may fail; remove the references or render matching feedback elements.
#8616
2026-10-20 17:27
cubic
inline · P2
frontend/src/pages/secret-manager/integrations/OctopusDeployConfigurePage/OctopusDeployConfigurePage.tsx:194defect-functional
Pressing Enter while a selector has focus never submits this form because the V3 `Combobox` prevents the default action even when its popup is closed. Preserve native Enter submission when no option is being committed.
#8616
2026-10-20 17:27
cubic
inline · P2
frontend/src/pages/secret-manager/integrations/GcpSecretManagerConfigurePage/GcpSecretManagerConfigurePage.tsx:430defect-functional
Change this field label to “Label Value”; both GCP label inputs are currently announced as “Label Name,” so users and screen readers cannot distinguish them. ```suggestion Label Value ```
#8616
2026-10-20 17:27
cubic
inline · P3
frontend/src/pages/secret-manager/integrations/NorthflankAuthorizePage/NorthflankAuthorizePage.tsx:67defect-functional
The input references an error element that is absent until validation fails, leaving a dangling `aria-describedby` IDREF in the normal state. Set this attribute only while the matching `FieldError` is rendered. ```suggestion aria-describedby={ apiKeyErrorText !== "" ? "NorthflankAuthorizePage-field-1-feedback" : undefined } ```
#8616
2026-10-20 17:27
cubic
inline · P3
frontend/src/pages/secret-manager/integrations/QoveryConfigurePage/QoveryConfigurePage.tsx:237defect-functional
These controls reference feedback IDs that are never rendered, leaving invalid ARIA ID references. Remove the `aria-describedby` attributes or render matching feedback elements.
#8616
2026-10-20 17:27
cubic
inline · P3
frontend/src/pages/secret-manager/integrations/QoveryConfigurePage/QoveryConfigurePage.tsx:209defect-functional
Replace “Checkly” with “Qovery”; this description tells users that the Qovery integration syncs to the wrong provider. ```suggestion Choose which environment in Infisical you want to sync to Qovery environment variables. ```
#8616
2026-10-20 17:34
greptile
issue-comment
—summary
Confidence Score: 5/5 **[Medium impact]** No blocking issue was identified; the change is safe to merge within the reviewed scope. Summary Migrates 17 native-integration pages from V2 to V3 components while keeping their requests, payloads, and success navigation. - Provider authorization pages use the shared V3 card layout. - Provider configuration pages use V3 fields and a shared card layout. - GCP’s overwrite warning uses the V3 dialog and action order. Intentional limitations acknowledged by PierreOucif: existing direct-URL pages remain; native `required` attributes are removed under G3; inherited `SecretPathInput` ref/blur limits remain; existing omitted select refs are not repaire
… the rest (343 characters)d. PierreOucif described provider-backed rendering and GCP dialog interactions as source-only checks. The additional GCP confirmation checks used an isolated rendered page with controlled dependencies, not a live provider connection. Reviews (1) · Last reviewed commit: "improvement(integrations): port native p..." · Reviewed by the reviewer
#8618
2026-10-20 17:36
coderabbit
issue-comment
—summary
No actionable comments were generated in the recent review. 🎉 📝 Walkthrough Authorization and configuration pages for seven integrations now use v3 UI components and mutation-provided pending states. The authentication-selection page also uses v3 card and button components. Existing integration fields, validation, and request flows remain represented. **Changes: Integration page UI** |Layer / File(s)|Summary| |:---|:---| |**Authentication selection layout** `frontend/src/pages/secret-manager/integrations/SelectIntegrationAuthPage/SelectIntegrationAuthPage.tsx`|The connection-selection card uses v3 card sections, updated connection buttons, and a v3 field separator.| |**Authorization fo
… the rest (2765 characters)rms and pending states** `frontend/src/pages/secret-manager/integrations/*AuthorizePage/*AuthorizePage.tsx`|Seven authorization forms use v3 fields, card sections, validation feedback, and documentation badges. Their buttons use pending state from the save mutation hook.| |**Configuration forms and pending states** `frontend/src/pages/secret-manager/integrations/*ConfigurePage/*ConfigurePage.tsx`|Eight configuration forms use v3 fields, selects, and card sections. Their buttons use pending state from the create mutation hook, while existing fields and disabled conditions remain. | **Priority:** ➖ Normal **Estimated code review effort:** 3 (Moderate) | ~25 minutes **Change:** Refactor **Suggested reviewers:** `andrewhuhh` Merge Risk: ⚪ Minimal · up to b08de No actionable PR-introduced issue remains; the change is mergeable after normal checks. Pre-merge checks: 5 | Check name | Status | Explanation | | :------------------------ | :--------------------------------------------------------------------------------------------: | :-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | Title check | | The title clearly and concisely describes the main change: migrating native provider creation pages to V3. It follows the conventional commit format. | | Description check | | The description is complete and follows the repository template. It explains the context, UI changes, verification steps, change type, checklist status, screenshots, preserved behavior, and known veri… | | Docstring Coverage | | No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check. Docstring coverage is scoped to functions touched by this diff. Analyzed 0 functions across 1… | | Linked Issues check | | Check skipped because no linked issues were found for this pull request. | | Out of Scope Changes check | | Check skipped because no linked issues were found for this pull request. | Full details: Description check **Explanation** The description is complete and follows the repository template. It explains the context, UI changes, verification steps, change type, checklist status, screenshots, preserved behavior, and known verification limitation. Full details: Docstring Coverage **Explanation** No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check. Docstring coverage is scoped to functions touched by this diff. Analyzed 0 functions across 16 files. ✨ Finishing Touches 📝 Generate docstrings 🧪 Generate unit tests (beta) Comment `@the reviewer help` to get the list of available commands.
#8618
2026-10-20 17:44
cubic
inline · P3
frontend/src/pages/secret-manager/integrations/WindmillConfigurePage/WindmillConfigurePage.tsx:4defect-functional
All three controls reference `*-feedback` IDs that this component never renders, leaving dangling ARIA descriptions. Remove these references or add matching feedback elements so assistive technology receives valid descriptions.
#8618
2026-10-20 17:44
cubic
review-body
—summary
**1 issue found** across 16 files Reply with feedback, questions, or to request a fix.View guided diff | Turn on auto-fix | Re-trigger the reviewer
#8618
2026-10-20 17:51
greptile
issue-comment
—summary
Confidence Score: 5/5 **[Medium impact]** No blocking findings remain; the reviewed changes are suitable to merge within the stated provider-page scope. Summary Migrates 16 native-integration pages from V2 to V3 components. - Provider connection pages now use V3 cards and fields. - Provider setup pages now use V3 cards and fields. - Saved and new connections appear as separate choices in a V3 card. No actionable issues were found. Source review and focused browser checks support merging within the stated provider-page scope. Browser coverage used actual page components with dummy local data and controlled responses; it did not exercise authenticated backend flows or real providers. Ack
… the rest (519 characters)nowledged scope from `PierreOucif`: direct-URL pages remain intentionally available until native integrations retire on August 19, 2027; existing copy and the Select Connection empty-list OAuth redirect are intentionally unchanged. Shared path-input ref forwarding and blur limitations are acknowledged as existing behavior and intentionally unchanged; no shared path-input fix is required for this migration. Reviews (1) · Last reviewed commit: "improvement(integrations): port native p..." · Reviewed by the reviewer
#8620
2026-10-20 17:53
coderabbit
issue-comment
—summary
No actionable comments were generated in the recent review. 🎉 📝 Walkthrough MFA confirmation now manages loading state around login completion. The MFA success callback returns the organization-selection promise when stored organization context exists, and returns `undefined` when it does not. **Changes: MFA completion** |Layer / File(s)|Summary| |:---|:---| |**MFA completion callbacks** `frontend/src/components/auth/Mfa.tsx`, `frontend/src/pages/auth/SelectOrgPage/SelectOrgPage.tsx`|The confirmation handler sets loading state before `completeLogin` and clears it in a `finally` block. The MFA success callback returns the organization-selection result when stored MFA organization contex
… the rest (3889 characters)t exists; otherwise, it returns `undefined`.| **Priority:** ⬇️ Low **Estimated code review effort:** 2 (Simple) | ~10 minutes **Change:** Bug fix **Suggested reviewers:** `andrewhuhh` Merge Risk: ⚪ Minimal · up to 7a9aa The MFA confirmation retry flow has no established merge-blocking risk. Security architecture risk: ⚪ Minimal · up to 7a9aa The change restores retry after organization-selection failure without repeating MFA verification or replacing recovery codes. The inspected flow retains server-side organization and MFA checks; no material security risk was found to be introduced or worsened. **Retained concerns** No architecture-level concerns identified. Security review details **Security Blast Radius** - _inferred_ — The changed behavior concerns the current user's completion flow for a selected root or sub-organization. Restoring the retry control does not expand independently attackable organization scope because client-supplied organization IDs remain subject to server-side membership checks on each request. **Security Findings and Attack Paths** - _inferred_ — No introduced MFA or organization-authorization bypass was identified in the inspected path. Bypassing the UI pending state still reaches the existing authenticated selection endpoint; returning its promise does not itself grant authority or mint credentials. **Trust Boundaries and Controls** - _observed_ — MFA assurance originates in a server-issued verified session token. Organization selection re-evaluates organization and MFA-method changes when deciding whether another challenge is required; the frontend loading flag is not the authorization control. Pre-merge checks: 5 | Check name | Status | Explanation | | :------------------------ | :--------------------------------------------------------------------------------------------: | :-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | Title check | | The title uses the conventional commit format and clearly describes the main change: allowing users to retry after MFA organization selection fails. | | Description check | | The description is complete and matches the repository template. It explains the problem and solution, provides verification steps, identifies the change as a fix, documents testing and limitations, a… | | Docstring Coverage | | No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check. Docstring coverage is scoped to functions touched by this diff. Analyzed 0 functions across 2… | | Linked Issues check | | Check skipped because no linked issues were found for this pull request. | | Out of Scope Changes check | | Check skipped because no linked issues were found for this pull request. | Full details: Description check **Explanation** The description is complete and matches the repository template. It explains the problem and solution, provides verification steps, identifies the change as a fix, documents testing and limitations, and completes the checklist. Full details: Docstring Coverage **Explanation** No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check. Docstring coverage is scoped to functions touched by this diff. Analyzed 0 functions across 2 files. ✨ Finishing Touches 📝 Generate docstrings 🧪 Generate unit tests (beta) ### Usage-based review receipt - Mode: Continue automatically - Reviewed files: 2 - Waived: $0.50 (charged $0.00) - View usage details > [!NOTE] > This review exceeded your plan’s limits and used usage-based reviews—free during trial, billed after paid activation unless disabled. **Manage usage-based reviews**. Comment `@the reviewer help` to get the list of available commands.
#8620
2026-10-20 17:53
cubic
review-body
—summary
**No issues found** across 2 files View guided diff | Turn on auto-fix | Re-trigger the reviewer
#8620
2026-10-20 18:06
greptile
issue-comment
—summary
Confidence Score: 5/5 **[Medium impact]** Safe to merge within the verified scope; no actionable defects were found in the changed confirmation and selection callbacks. Summary This PR lets users retry organization selection after first-time MFA without repeating verification. - MFA users can retry organization selection after it fails. No blocking issues were found; the change is safe to merge within the verified scope. These checks used the real recovery components with controlled callbacks. Authenticated first-time MFA against a licensed backend remains unverified. Reviews (1) · Last reviewed commit: "fix(auth): release recovery-code confirm..." · Reviewed by the reviewer
#8621
2026-10-20 18:08
coderabbit
issue-comment
—summary
No actionable comments were generated in the recent review. 🎉 📝 Walkthrough Combobox creation now supports dialog, sheet, and action modes. The single- and multi-select flows share form overlay handling. New Storybook examples demonstrate the creation workflows, and the migration guide documents their contracts. **Changes: Combobox creation modes** |Layer / File(s)|Summary| |:---|:---| |**Creation contracts and shared overlays** `frontend/src/components/v3/generic/Combobox/Combobox.tsx`, `frontend/src/components/v3/generic/Combobox/index.ts`, `frontend/src/components/V3_MIGRATION.md`|Added shared form render props and sheet and action configuration types. Dialog and sheet forms use sha
… the rest (5520 characters)red overlay state and dismissal handling. Updated the public exports and migration guidance.| |**Selection flows and creation examples** `frontend/src/components/v3/generic/Combobox/Combobox.tsx`, `frontend/src/components/v3/generic/Combobox/Combobox.creation.stories.tsx`|Single- and multi-select flows route action creation through `onAction` and open shared form overlays for dialog and sheet modes. Storybook examples demonstrate sheet, dialog, and action workflows.| **Priority:** ⬇️ Low **Estimated code review effort:** 3 (Moderate) | ~25 minutes **Change:** Feature ### Sequence Diagram(s) ```mermaid sequenceDiagram participant User participant Combobox participant CreationOverlay participant Form participant Caller User->>Combobox: Select a form creation option Combobox->>CreationOverlay: Open dialog or sheet CreationOverlay->>Form: Render with initialInput, complete, and cancel Form->>Caller: Submit form data Caller->>Combobox: Update options and selection, then complete ``` **Suggested reviewers:** `andrewhuhh` Merge Risk: ⚪ Minimal · up to 2d012 This change adds Sheet-based and caller-owned action creation modes to the Combobox, along with Storybook examples and migration guidance. Existing inline and dialog behavior is reported unchanged. No concrete defects were found, so the change is ready to merge. The authenticated PKI creation and policy sub-wizard flows are reported as untested and are worth checking when the PKI callers adopt these modes. Security architecture risk: 🔵 Low · up to 2d012 The change is limited to shared UI creation workflows. Persistence and selection remain caller-owned, and the inspected dispatch paths do not establish a new security vulnerability. Action workflows must supply their own validation and pending-work controls; production integrations were not verified. **Retained concerns** No architecture-level concerns identified. Security review details **Security Blast Radius** - _inferred_ — The demonstrated exposure is a shared frontend callback contract carrying user-entered query text. Tenant, asset, service, or datastore consequences depend on downstream integrations; no such privileged integration was verified in the inspected scope. **Trust Boundaries and Controls** - _observed_ — Action dispatch checks isDisabled in both selection variants. This is a UI invocation control, not a demonstrated authorization boundary. Action intentionally delegates validation, pending work, and eventual persistence to its caller. - _observed_ — Form opening rejects disabled or pending configurations, cancellation rejects pending work, and dismissal can pass through caller-owned discard interception. Sheet receives the same pending state and connects existing dismissal guards, disabled close behavior, and focus-return callbacks. **Resilience and Maintainability Implications** - _observed_ — Completion callbacks lack session identity and at-most-once guards, so retained callbacks can close an overlay after cancellation or reopening. This behavior predates the PR in dialog creation and is reused for Sheet. The inspected completion sinks affect search, popup state, and focus rather than persisting or selecting domain objects; a worsened security outcome was not established. Pre-merge checks: 5 | Check name | Status | Explanation | | :------------------------ | :--------------------------------------------------------------------------------------------: | :-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | Description check | | The description is complete and aligned with the template. It explains the context, documents UI changes with screenshots, lists verification steps, identifies the change type, and includes the checkl… | | Title check | | The title clearly and concisely describes the main change: adding Sheet and in-place Combobox creation support. It follows the conventional commit format. | | Docstring Coverage | | No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check. Docstring coverage is scoped to functions touched by this diff. Analyzed 0 functions across 3… | | Linked Issues check | | Check skipped because no linked issues were found for this pull request. | | Out of Scope Changes check | | Check skipped because no linked issues were found for this pull request. | Full details: Description check **Explanation** The description is complete and aligned with the template. It explains the context, documents UI changes with screenshots, lists verification steps, identifies the change type, and includes the checklist. Full details: Docstring Coverage **Explanation** No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check. Docstring coverage is scoped to functions touched by this diff. Analyzed 0 functions across 3 files. (1 skipped: 1 unsupported.) ✨ Finishing Touches 📝 Generate docstrings 🧪 Generate unit tests (beta) ### Usage-based review receipt - Mode: Continue automatically - Reviewed files: 4 - Waived: $1.00 (charged $0.00) - View usage details > [!NOTE] > This review exceeded your plan’s limits and used usage-based reviews—free during trial, billed after paid activation unless disabled. **Manage usage-based reviews**. Comment `@the reviewer help` to get the list of available commands.
#8621
2026-10-20 18:15
cubic
review-body
—summary
**No issues found** across 4 files View guided diff | Turn on auto-fix | Re-trigger the reviewer
#8621
2026-10-20 18:21
greptile
issue-comment
—summary
Confidence Score: 5/5 **[Medium impact]** Safe to merge within the shared Combobox scope; no blocking issues were established. Summary Adds Sheet-backed creation forms and caller-owned, in-place creation actions to `Combobox`, along with composition stories, migration guidance, and public type exports. Both Sheet compositions passed rendered lifecycle checks: dismissing a nested popup left the creation Sheet open, Keep Editing preserved the draft, Discard restored the parent query and focus, and successful creation did not submit the parent attachment form. Action creation passed all 32 combinations of single/multiple selection, query type, pointer/keyboard activation, and enabled/disabl
… the rest (848 characters)ed state. Enabled callbacks received the expected trimmed query, and disabled actions did not invoke them. The established protections remain intact: pending saves block Sheet dismissal, and starting an action preserves selected options. No blocking issues were found. This change is safe to merge within its shared-component scope. Scope acknowledgment: PierreOucif explicitly notes that no new test suite was committed and that authenticated PKI creation and the policy sub-wizard remain outside this change. The extracted real form and domain workflows still require validation before that migration can be considered ready. - Combobox creation forms can open in a Sheet. - The Create row can start a caller-owned workflow in place. Reviews (1) · Last reviewed commit: "improvement(ui): support Sheet and in-pl..." · Reviewed by the reviewer
#8623
2026-10-20 18:23
coderabbit
issue-comment
—summary
No actionable comments were generated in the recent review. 🎉 📝 Walkthrough Azure ADCS and Venafi signing-configuration forms now use independently scrollable content containers with separately positioned footers. Their fields and submission behavior remain unchanged. **Changes: Signing form layout** |Layer / File(s)|Summary| |:---|:---| |**Scrollable signing forms** `frontend/src/pages/cert-manager/CertAuthDetailsByIDPage/components/CaSigningConfigSection.tsx`|Azure ADCS and Venafi fields are in padded, independently scrollable containers. Each form’s footer sits outside its scrollable area. Field and submission behavior remain unchanged.| **Priority:** ⬇️ Low **Estimated code revie
… the rest (2102 characters)w effort:** 2 (Simple) | ~10 minutes **Change:** Bug fix **Suggested reviewers:** `andrewhuhh` Merge Risk: ⚪ Minimal · up to 9a6db The layout change has no verified merge-blocking risk in the supplied evidence. Pre-merge checks: 5 | Check name | Status | Explanation | | :------------------------ | :--------------------------------------------------------------------------------------------: | :-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | Title check | | The title clearly and concisely describes the primary change: anchoring signing configuration sheet footers. It follows the conventional commit format. | | Description check | | The description is complete and relevant. It explains the prior and new behavior, includes UI screenshots, verification steps, change type, checklist status, test results, and known verification limit… | | Docstring Coverage | | No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check. Docstring coverage is scoped to functions touched by this diff. Analyzed 0 functions across 1… | | Linked Issues check | | Check skipped because no linked issues were found for this pull request. | | Out of Scope Changes check | | Check skipped because no linked issues were found for this pull request. | Full details: Description check **Explanation** The description is complete and relevant. It explains the prior and new behavior, includes UI screenshots, verification steps, change type, checklist status, test results, and known verification limits. Full details: Docstring Coverage **Explanation** No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check. Docstring coverage is scoped to functions touched by this diff. Analyzed 0 functions across 1 files. ✨ Finishing Touches 📝 Generate docstrings 🧪 Generate unit tests (beta) Comment `@the reviewer help` to get the list of available commands.
#8623
2026-10-20 18:28
cubic
review-body
—summary
**No issues found** across 1 file View guided diff | Turn on auto-fix | Re-trigger the reviewer
#8623
2026-10-20 18:32
greptile
issue-comment
—summary
Confidence Score: 5/5 **[Low impact]** Safe to merge. Both signing forms retain visible action footers, reachable fields, and working native form submission in the tested short viewports. Summary Moves scrolling into a separate body inside the Azure ADCS and Venafi signing forms, keeping Cancel and Save outside the scrolling body and inside the same form. Browser checks of both providers at 1000×320 and 600×240 confirmed that the footer remains visible while the body scrolls, all fields remain reachable, and Save submits the correct form. These checks mounted the real changed component and shared UI primitives with synthetic domain data and persistence callbacks; connected-provider behav
… the rest (640 characters)ior and authenticated authorization acceptance were not tested. Focused lint and frontend typechecking passed. PierreOucif explicitly identified the footer alignment change as intentional: removing the negative-margin/alignment compensation restores default SheetFooter styling. Submit handlers, field values, reset logic, pending guards, and Native ADCS remain unchanged. No actionable issues were found; the scoped layout change is safe to merge. - Azure ADCS and Venafi sheets keep their actions visible while fields scroll. Reviews (1) · Last reviewed commit: "fix(cert-manager): anchor signing config..." · Reviewed by the reviewer
#8624
2026-10-20 18:34
coderabbit
issue-comment
—summary
No actionable comments were generated in the recent review. 🎉 📝 Walkthrough The identity modal now uses `DialogBody` and `DialogFooter` to structure its form. The name input no longer uses autofocus. Existing role, delete-protection, and metadata controls retain their behavior. **Changes: Identity modal layout** |Layer / File(s)|Summary| |:---|:---| |**Structure the identity modal form** `frontend/src/pages/organization/AccessManagementPage/components/OrgIdentityTab/components/IdentitySection/OrgIdentityModal.tsx`|The form fields are placed in `DialogBody`, and the action buttons are placed in `DialogFooter`. The name input no longer uses autofocus. Existing control behavior remains un
… the rest (2015 characters)changed.| **Priority:** ⬇️ Low **Estimated code review effort:** 2 (Simple) | ~10 minutes **Change:** Bug fix **Suggested reviewers:** `andrewhuhh` Merge Risk: ⚪ Minimal · up to e672e The identity modal now keeps its action buttons in a footer and scrolls its fields in the body, and the name field no longer takes focus automatically. No concrete merge-blocking risk was found. Normal runtime verification of the update flow is still advisable. Pre-merge checks: 5 | Check name | Status | Explanation | | :------------------------ | :--------------------------------------------------------------------------------------------: | :-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | Title check | | The title clearly identifies the UI change and follows the conventional commit format. | | Description check | | The description includes context, screenshots, verification steps, change type, checklist status, test results, and known limitations. It clearly identifies the in-scope and out-of-scope behavior. | | Docstring Coverage | | No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check. Docstring coverage is scoped to functions touched by this diff. Analyzed 0 functions across 1… | | Linked Issues check | | Check skipped because no linked issues were found for this pull request. | | Out of Scope Changes check | | Check skipped because no linked issues were found for this pull request. | Full details: Docstring Coverage **Explanation** No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check. Docstring coverage is scoped to functions touched by this diff. Analyzed 0 functions across 1 files. ✨ Finishing Touches 📝 Generate docstrings 🧪 Generate unit tests (beta) Comment `@the reviewer help` to get the list of available commands.
#8624
2026-10-20 18:40
cubic
inline · P3
frontend/src/pages/organization/AccessManagementPage/components/OrgIdentityTab/components/IdentitySection/OrgIdentityModal.tsx:152defect-functional
The `DialogBody` now has `overflow-y-auto` and no padding, so the 3px focus ring on the full-width inputs is clipped at the top, left, and right edges. Add a small inset padding to the body so the ring stays visible.
#8624
2026-10-20 18:40
cubic
review-body
—summary
**1 issue found** across 1 file Reply with feedback, questions, or to request a fix.View guided diff | Turn on auto-fix | Re-trigger the reviewer
#8624
2026-10-20 18:44
greptile
issue-comment
—summary
Confidence Score: 5/5 **[Low impact]** The change is safe to merge within its stated composition-only scope; the tested rendering preserves contained scrolling and native form behavior. Summary The PR puts the identity edit fields inside DialogBody and the action buttons inside DialogFooter. Cancel now uses ghost and still precedes Update. Removing input autoFocus lets the dialog choose initial focus; submission, validation, reset, and license-check code remain unchanged. Browser checks using the actual child form, dialog primitives, application styles, and React Hook Form found no actionable regression in the tested composition. With 24 metadata rows at short viewport heights, the body
… the rest (1055 characters)scrolled while the header and footer remained visible. Cancel did not submit; Update and Enter submitted expected values; invalid fields received focus; and Update remained disabled during submission. These checks used deterministic organization, subscription, and role fixtures, an asynchronous update sink, and a closed UpgradeGate stub. They do not establish authenticated persistence or normal-runtime permission and license behavior. Acknowledged scope limits: PierreOucif explicitly identified the inherited role-picker Escape behavior as outside this change; it remains neither fixed nor accepted. PierreOucif also identified missing parent pending-dismiss locks as an inherited limitation, with pending-state propagation outside this bounded change. Neither is reported as a new defect or treated as fixed. No blocking issues were found in the reviewed change. - Machine identity fields scroll while the dialog actions stay together. Reviews (1) · Last reviewed commit: "fix(ui): compose machine identity dialog..." · Reviewed by the reviewer
#8625
2026-10-20 18:47
coderabbit
issue-comment
—summary
No actionable comments were generated in the recent review. 🎉 📝 Walkthrough PamDetailSheet now has independently scrollable sidebar and content regions with separate footer slots. PAM approval, access-request, and discovery sheets place their existing action controls in those slots. **Changes: PAM detail sheet layout** |Layer / File(s)|Summary| |:---|:---| |**Detail sheet structure and footer slots** `frontend/src/pages/pam/components/PamDetailSheet.tsx`|PamDetailSheet adds accessible sheet headers, separate scrolling regions, and dedicated sidebar and content footer props.| **PAM action placement** |Layer / File(s)|Summary| |:---|:---| |**Approval, access-request, and discovery acti
… the rest (3865 characters)ons** `frontend/src/pages/pam/PamApprovalRequestsPage/components/ApprovalRequestDetailSheet.tsx`, `frontend/src/pages/pam/components/RequestAccessSheet.tsx`, `frontend/src/pages/pam/PamDiscoveryPage/components/DiscoverySourceDetailSheet.tsx`|Approval and access-request controls move to the content footer. The discovery scan button moves to the sidebar footer. Existing action behavior is preserved.| **Priority:** ⬇️ Low **Estimated code review effort:** 2 (Simple) | ~12 minutes **Change:** Bug fix **Suggested reviewers:** `lb-vn` Merge Risk: ⚪ Minimal · up to ad73f This change anchors PAM detail sheet headers and action footers while keeping the existing action behavior. No concrete merge-blocking risk was identified. Security architecture risk: ⚪ Minimal · up to ad73f The change reorganizes existing PAM controls without introducing new privileges or changing their conditions, callbacks, or submitted data. The relocated submit button remains associated with its validated form. No material security risk was found to be introduced or worsened. **Retained concerns** No architecture-level concerns identified. Security review details **Security Blast Radius** - _inferred_ — The changed exposure is the placement and accessibility of controls within existing PAM sheets. The comparison does not add callers, credentials, resource targets, or mutation authority. Shared-shell layout affects its consumers, but privileged actions remain supplied by those consumers. **Trust Boundaries and Controls** - _observed_ — Existing frontend controls survive relocation: review actions remain status-gated, Approve remains disabled for the requester's own request, revoke retains confirmation, Scan Now retains missing-source and mutation-pending controls, and request submission remains absent in the business-pending view. These observations do not independently verify server-side authorization. **Resilience and Maintainability Implications** - _observed_ — PamDetailSheet still does not pass mutation-pending state into SheetContent, so sheet-level pending-dismiss locking remains absent. This condition predates the PR; the comparison preserves existing dismissal and mutation paths rather than establishing cancellation or stronger concurrency guarantees. Pre-merge checks: 5 | Check name | Status | Explanation | | :------------------------ | :--------------------------------------------------------------------------------------------: | :-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | Title check | | The title uses the conventional commit format and clearly summarizes the main change: anchoring PAM detail-sheet headers and action footers. | | Description check | | The description includes the required context, screenshots, verification steps, type selection, and checklist. It also documents testing results and known limitations. | | Docstring Coverage | | No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check. Docstring coverage is scoped to functions touched by this diff. Analyzed 0 functions across 4… | | Linked Issues check | | Check skipped because no linked issues were found for this pull request. | | Out of Scope Changes check | | Check skipped because no linked issues were found for this pull request. | Full details: Docstring Coverage **Explanation** No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check. Docstring coverage is scoped to functions touched by this diff. Analyzed 0 functions across 4 files. ✨ Finishing Touches 📝 Generate docstrings 🧪 Generate unit tests (beta) Comment `@the reviewer help` to get the list of available commands.
#8625
2026-10-20 18:54
cubic
inline · P3
frontend/src/pages/pam/PamApprovalRequestsPage/components/ApprovalRequestDetailSheet.tsx:147defect-functional
These actions now align to the left because `SheetFooter` does not apply `justify-end`, while the removed row did. Preserve the existing right alignment by adding a full-width, right-aligned wrapper or applying `justify-end` to the footer.
#8625
2026-10-20 18:54
cubic
inline · P3
frontend/src/pages/pam/components/PamDetailSheet.tsx:156defect-functional
The sidebar title loses its previous `text-lg` styling because `SheetTitle` supplies no font size. Add `text-lg` to preserve the existing title hierarchy. ```suggestion <SheetTitle className="text-lg">{title || "PAM Details"}</SheetTitle> ```
#8625
2026-10-20 18:54
cubic
review-body
—summary
**2 issues found** across 4 files Reply with feedback, questions, or to request a fix.View guided diff | Turn on auto-fix | Re-trigger the reviewer
#8625
2026-10-20 18:58
greptile
issue-comment
—summary
Confidence Score: 5/5 **[Medium impact]** No blocking issue was identified in the changed composition; authenticated product behavior remains unverified. Summary Updates PAM detail sheets to keep headers and action footers separate from scrolling content. - PAM detail sheets separate headings and actions from scrolling details. - Access request buttons stay available outside the scrolling form. Browser checks using source-attributed components and real UI primitives confirmed reachable scrolling content and visible footer actions at normal and short viewport heights, including long wrapped titles. Accessible dialog titles and optional descriptions behaved correctly. Access-request checks
… the rest (1057 characters) confirmed required-field validation and focus, textarea newline behavior, exactly-once submission, distinct form IDs, and Cancel behavior. These checks used bounded synthetic renders, not an authenticated product session. API and metadata boundaries were stubbed for access-request submission; authenticated mutations and permissions were not verified. Acknowledged limits from PierreOucif: nested Select Escape behavior remains unresolved and is not fixed or accepted here; pending-dismiss locks remain absent because mutation pending state is not propagated into SheetContent; authenticated mutation and permission checks remain outside the supplied verification. PierreOucif described these as inherited limits outside this composition change, with picker/primitives, save-bar tab editors, and pending-state propagation outside its bounded scope. No actionable issues were found. No blocking issue was identified in the changed composition. Reviews (1) · Last reviewed commit: "fix(pam): anchor detail sheet headers an..." · Reviewed by the reviewer
#8626
2026-10-20 19:02
coderabbit
issue-comment
—summary
📝 Walkthrough Several forms now use `SheetFooter` for their action areas. The dynamic-secret form header sits outside the scrollable content, and the selected-provider header uses default styling. **Changes: Sheet Form Layout** |Layer / File(s)|Summary| |:---|:---| |**SheetFooter adoption** `frontend/src/components/dynamic-secrets/DynamicSecretProviderForm/DynamicSecretProviderForm.tsx`, `frontend/src/components/pki-syncs/forms/*`, `frontend/src/components/proxied-services/forms/*`, `frontend/src/components/secret-rotations-v2/forms/SecretRotationV2Form.tsx`, `frontend/src/components/secret-syncs/forms/*`|These forms use `SheetFooter` in place of custom footer wrappers. The PKI sync and
… the rest (2451 characters)secret sync edit forms place Save with the right-side navigation controls.| |**Dynamic-secret header layout** `frontend/src/components/dynamic-secrets/DynamicSecretProviderForm/DynamicSecretProviderForm.tsx`, `frontend/src/pages/secret-manager/OverviewPage/components/CreateDynamicSecretForm/CreateDynamicSecretForm.tsx`|The dynamic-secret form header is outside the scrollable content. The selected-provider `SheetHeader` no longer uses negative margins.| **Priority:** ⬇️ Low **Estimated code review effort:** 2 (Simple) | ~10 minutes **Change:** Bug fix **Suggested reviewers:** `andrewhuhh` Merge Risk: 🔵 Low · up to 7b604 On sufficiently narrow screens, dynamic-secret users may be unable to reach the rightmost footer action; restore responsive stacking before or immediately after merge. Pre-merge checks: 5 | Check name | Status | Explanation | | :------------------------ | :--------------------------------------------------------------------------------------------: | :-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | Title check | | The title uses conventional commit format and clearly identifies the primary UI change: replacing custom footer markup with canonical Sheet footers in reusable wizard forms. | | Description check | | The description includes the required Context, Screenshots, Steps to verify the change, Type, and Checklist sections. It explains the scope, verification performed, and known verification limits. | | Docstring Coverage | | No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check. Docstring coverage is scoped to functions touched by this diff. Analyzed 0 functions across 9… | | Linked Issues check | | Check skipped because no linked issues were found for this pull request. | | Out of Scope Changes check | | Check skipped because no linked issues were found for this pull request. | Full details: Docstring Coverage **Explanation** No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check. Docstring coverage is scoped to functions touched by this diff. Analyzed 0 functions across 9 files. ✨ Finishing Touches 📝 Generate docstrings 🧪 Generate unit tests (beta) Comment `@the reviewer help` to get the list of available commands.
#8626
2026-10-20 19:04
cubic
inline · P2
frontend/src/components/dynamic-secrets/DynamicSecretProviderForm/DynamicSecretProviderForm.tsx:460defect-functional
This footer now keeps Back, Cancel, and the provider submit button in one non-wrapping row on narrow sheets. Restore the responsive column layout so long labels do not overflow or compress the primary action. ```suggestion <SheetFooter className="shrink-0 flex-col-reverse items-stretch border-t @sm:flex-row @sm:items-center @sm:justify-end"> ```
#8626
2026-10-20 19:04
cubic
review-body
—summary
**1 issue found** across 9 files Reply with feedback, questions, or to request a fix.View guided diff | Turn on auto-fix | Re-trigger the reviewer
#8626
2026-10-20 19:10
coderabbit
inline · P2
frontend/src/components/dynamic-secrets/DynamicSecretProviderForm/DynamicSecretProviderForm.tsx:477defect-functional
**Restore a stacked footer at narrow widths.** `CreateDynamicSecretForm` reaches this footer with Back, Cancel, and Submit. The form Sheet is full width at narrow viewports, but the footer is a non-wrapping row. Each button is `shrink-0` and `whitespace-nowrap`, so the row can exceed the available width. The rightmost action can then extend outside the Sheet viewport. Suggested fix ```diff --- "a/frontend/src/components/dynamic-secrets/DynamicSecretProviderForm/DynamicSecretProviderForm.tsx" +++ "b/frontend/src/components/dynamic-secrets/DynamicSecretProviderForm/DynamicSecretProviderForm.tsx" @@ -457,11 +457,17 @@ </section> </div> - <SheetFooter className="s
… the rest (1747 characters)hrink-0 border-t"> + <SheetFooter className="shrink-0 border-t flex-col-reverse @sm:flex-row @sm:justify-end"> {onBack && ( - <Button type="button" variant="ghost" onClick={onBack} isDisabled={isPending}> + <Button + type="button" + variant="ghost" + onClick={onBack} + isDisabled={isPending} + className="@sm:mr-auto" + > Back </Button> )} <Button type="button" variant="ghost" onClick={onCancel} isDisabled={isPending}> Cancel ``` 📝 Committable suggestion > ‼️ **IMPORTANT** > Carefully review the code before committing. Ensure that it accurately replaces the highlighted code, contains no missing lines, and has no issues with indentation. Thoroughly test & benchmark the code to ensure it meets the requirements. ```suggestion <SheetFooter className="shrink-0 border-t flex-col-reverse @sm:flex-row @sm:justify-end"> {onBack && ( <Button type="button" variant="ghost" onClick={onBack} isDisabled={isPending} className="@sm:mr-auto" > Back </Button> )} <Button type="button" variant="ghost" onClick={onCancel} isDisabled={isPending}> Cancel </Button> <Button type="submit" variant="project" isPending={isPending} isDisabled={providerSubmitState.isDisabled || isPending} > {mode === "create" ? definition.create.submitLabel : definition.edit.submitLabel} </Button> </SheetFooter> ```
#8626
2026-10-20 19:10
coderabbit
review-body
—summary
**Actionable comments posted: 1**
#8626
2026-10-20 19:12
greptile
issue-comment
—summary
Confidence Score: 5/5 **[Low impact]** No identified defect blocks merging; source inspection supports preserved form ownership and button behavior, but rendered acceptance remains incomplete. Summary Replaces custom footer containers with `SheetFooter` across eight reusable forms. - Reusable wizard forms keep their actions in the shared sheet footer. - Edit sync forms put Save after Cancel and step navigation. - Dynamic-secret setup keeps its provider header above the scrolling fields. No actionable issues were established. Source inspection supports the composition-only change, and no identified defect blocks merging. Rendered behavior and provider submissions remain unverified. Ackno
… the rest (877 characters)wledged limitation: PierreOucif explicitly preserves dismissal while a request is pending because the outer hosts do not pass `SheetContent.isPending`. This existing limitation is not reported as a new issue. ### the check validation blocked The dynamic-secret browser check encountered a React-refresh initialization error and exhausted its execution budget before the correction could be retested. The edit-sync browser check could not render because its synthetic harness omitted the `GcpCertificateManagerScope` export required by the PKI schema. No keyboard, submission, or short-height layout assertions completed. Evidence uploads were unavailable because no artifact-upload operation, authenticated upload endpoint, or documented upload command was supplied. Reviews (1) · Last reviewed commit: "fix(ui): use canonical Sheet footers in ..." · Reviewed by the reviewer
#8627
2026-10-20 19:14
coderabbit
issue-comment
—summary
No actionable comments were generated in the recent review. 🎉 📝 Walkthrough Nine certificate manager views now use the shared `SheetFooter` component instead of manually styled footer wrappers. Existing controls and actions remain in place. **Changes: Certificate manager footer adoption** |Layer / File(s)|Summary| |:---|:---| |**Replace manual footer wrappers** `frontend/src/pages/cert-manager/CertificateAuthoritiesPage/components/CreateCaWizard/CreateCaWizard.tsx`, `frontend/src/pages/cert-manager/CertificateAuthoritiesPage/components/ExternalCaModal.tsx`, `frontend/src/pages/cert-manager/CodeSigningPage/components/CreateSignerWizard/CreateSignerWizard.tsx`, `frontend/src/pages/cert-m
… the rest (2980 characters)anager/PoliciesPage/components/CertificatePoliciesTab/CreatePolicyModal.tsx`, `frontend/src/pages/cert-manager/SettingsPage/components/HsmConnectorsTab/CreateHsmConnectorWizard/CreateHsmConnectorWizard.tsx`, `frontend/src/pages/cert-manager/SettingsPage/components/HsmConnectorsTab/EditHsmConnectorSheet.tsx`, `frontend/src/pages/cert-manager/SignerDetailPage/components/EditSignerModal/EditSignerModal.tsx`, `frontend/src/pages/cert-manager/SignerDetailPage/components/EditSignerPolicyModal/EditSignerPolicyModal.tsx`, `frontend/src/pages/cert-manager/components/CertificateAlerts/CertificateAlertSheet.tsx`|These views replace manual footer wrappers with `SheetFooter`. Their existing navigation, form actions, indicators, and related footer content remain in place.| **Priority:** ⬇️ Low **Estimated code review effort:** 2 (Simple) | ~10 minutes **Change:** Refactor **Suggested reviewers:** `carlosmonastyrski` Merge Risk: ⚪ Minimal · up to c80d1 This is a localized footer-wrapper adoption with no established behavioral regression, so it is mergeable. Pre-merge checks: 5 | Check name | Status | Explanation | | :------------------------ | :--------------------------------------------------------------------------------------------: | :-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | Title check | | The title clearly and concisely describes the main change: replacing manual footers with canonical Sheet footers in Certificate Manager. | | Description check | | The description includes all required sections, explains the scope and behavior, provides screenshots and verification steps, identifies the change type, and completes the checklist. | | Docstring Coverage | | No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check. Docstring coverage is scoped to functions touched by this diff. Analyzed 0 functions across 9… | | Linked Issues check | | Check skipped because no linked issues were found for this pull request. | | Out of Scope Changes check | | Check skipped because no linked issues were found for this pull request. | Full details: Docstring Coverage **Explanation** No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check. Docstring coverage is scoped to functions touched by this diff. Analyzed 0 functions across 9 files. ✨ Finishing Touches 📝 Generate docstrings 🧪 Generate unit tests (beta) ### Usage-based review receipt - Mode: Continue automatically - Reviewed files: 9 - Waived: $2.25 (charged $0.00) - View usage details > [!NOTE] > This review exceeded your plan’s limits and used usage-based reviews—free during trial, billed after paid activation unless disabled. **Manage usage-based reviews**. Comment `@the reviewer help` to get the list of available commands.
#8627
2026-10-20 19:15
cubic
inline · P2
frontend/src/pages/cert-manager/CertificateAuthoritiesPage/components/ExternalCaModal.tsx:1138defect-functional
This footer loses its original `px-6` and `gap-3`, so the buttons sit 16px from the edge instead of 24px and the gap shrinks to 8px. Add `px-6 gap-3` to the className to keep the existing layout, as CertificateWizardSheet does. ```suggestion <SheetFooter className="shrink-0 items-center justify-end gap-3 border-t px-6"> ```
#8627
2026-10-20 19:15
cubic
inline · P2
frontend/src/pages/cert-manager/CodeSigningPage/components/CreateSignerWizard/CreateSignerWizard.tsx:474defect-functional
The footer's original `px-6 py-4` and `gap-3` were dropped in this swap, and `SheetFooter` defaults to `p-4 gap-2`. Footer padding shrinks from 24px to 16px horizontally, and the gap between the status text and the buttons shrinks too. Restore the original spacing classes on the `SheetFooter` so the footer keeps its layout. ```suggestion <SheetFooter className="shrink-0 items-center justify-between gap-3 border-t px-6 py-4"> ```
#8627
2026-10-20 19:15
cubic
inline · P2
frontend/src/pages/cert-manager/SignerDetailPage/components/EditSignerPolicyModal/EditSignerPolicyModal.tsx:310defect-functional
`SheetFooter` brings its own `gap-2 p-4`, so the action row loses its original `gap-3` and `px-6` spacing. Restore those values in the `className` to keep the footer layout unchanged. ```suggestion <SheetFooter className="shrink-0 items-center justify-between gap-3 border-t px-6"> ```
#8627
2026-10-20 19:15
cubic
review-body
—summary
**4 issues found** across 9 files Reply with feedback, questions, or to request a fix.View guided diff | Turn on auto-fix | Re-trigger the reviewer
#8627
2026-10-20 19:15
cubic
inline · P3
frontend/src/pages/cert-manager/components/CertificateAlerts/CertificateAlertSheet.tsx:248defect-functional
The footer's horizontal padding drops from `px-6` to `p-4` and the button gap drops from `gap-3` to `gap-2`, because the `SheetFooter` defaults now apply. The PR description says the change preserves existing layout, so either note this spacing change or keep the original values by adding them to `className`. ```suggestion <SheetFooter className="shrink-0 items-center justify-between gap-3 border-t px-6 py-4"> ```
#8627
2026-10-20 19:24
greptile
issue-comment
—summary
Confidence Score: 5/5 **[Low impact]** No blocking findings identified; the footer substitutions preserve the existing form placement and action handlers, although independent browser checks remain incomplete. Summary Replaces manual footer wrappers with `SheetFooter` in nine Certificate Manager forms and wizards. - Certificate Manager forms now use the sheet’s shared footer. Acknowledged limits: - PierreOucif explicitly identifies CreateSigner’s compatibility CA `FilterableSelect` Escape behavior and the existing pending-dismiss/Back/Stepper gaps as unchanged and out of scope. - PierreOucif states that HSM Credentials/final-create cannot be reached without a PKCS#11 Gateway and does no
… the rest (589 characters)t claim HSM Verify, external issuance, signer save/reissue, alert delivery, OAuth, or simulated mutation-response acceptance. These capabilities are unchanged by the footer substitutions. ### the check validation blocked Independent browser checks could not complete because Vite could not resolve React from the isolated rendering harness. Short-viewport layout, native submission, Cancel, and read-only Close behavior were therefore not independently verified by these checks. Reviews (1) · Last reviewed commit: "fix(cert-manager): use canonical Sheet f..." · Reviewed by the reviewer
#8630
2026-10-20 19:26
coderabbit
issue-comment
—summary
📝 Walkthrough The changes remove the `NoticeBannerV2` component and its export, and remove the `AdminTable` and `AdminV3Adapters` modules. The lifecycle and migration records are updated to document the `NoticeBannerV2` removal. **Changes: NoticeBannerV2 removal** |Layer / File(s)|Summary| |:---|:---| |**Remove NoticeBannerV2 and update records** `frontend/src/components/v2/NoticeBannerV2/NoticeBannerV2.tsx`, `frontend/src/components/v2/index.tsx`, `frontend/src/components/COMPONENT_LIFECYCLE.md`, `frontend/src/components/V3_MIGRATION.md`|The component implementation and barrel export are removed. The lifecycle and migration records now document the removal.| **Admin component module re
… the rest (3794 characters)movals** |Layer / File(s)|Summary| |:---|:---| |**Remove admin table and adapter modules** `frontend/src/pages/admin/components/AdminTable.tsx`, `frontend/src/pages/admin/components/AdminV3Adapters.tsx`|The `AdminTable` module and the `AdminV3Adapters` module, including their exported components and aliases, are removed.| **Priority:** ⬇️ Low **Estimated code review effort:** 2 (Simple) | ~10 minutes **Change:** Other **Suggested reviewers:** `andrewhuhh` Merge Risk: 🔵 Low · up to a1fb4 The code removals have no identified consumers, but the migration ledger should be corrected before merge. Architecture risk: 🔵 Low · up to a1fb4 The change affects 1 system. **Changed systems:** `frontend` **Architecture concerns** No architecture-level concerns identified. Review details **Systems and components** - _observed_ — frontend (ui) was modified; 6 changed files map to changed impact. **Before / after behavior** - _observed_ — Modified behavior in frontend/src/components/COMPONENT_LIFECYCLE.md: `NoticeBannerV2` changed from blocked pending V3 `Alert` verification to removed, with its implementation and export surfaces documented as deleted after no consumers were found. - _observed_ — Modified behavior in frontend/src/components/V3_MIGRATION.md: The `NoticeBannerV2` coverage row now documents removal of the unused implementation, family index, and root export instead of planned migration to V3 Alert. - _observed_ — Modified behavior in frontend/src/components/v2/NoticeBannerV2/NoticeBannerV2.tsx: The `NoticeBannerV2` component and its imports, props type, styling, info icon, title rendering, and child-content rendering were removed. - _observed_ — Modified behavior in frontend/src/components/v2/index.tsx: The `NoticeBannerV2` re-export was removed from the component barrel file. Pre-merge checks: 5 | Check name | Status | Explanation | | :------------------------ | :--------------------------------------------------------------------------------------------: | :-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | Title check | | The title clearly and concisely describes the main changes: deleting the orphan NoticeBanner family and unused admin adapters. It follows the conventional commit format. | | Description check | | The description is complete and relevant. It explains the context, lists the removed sources, states that screenshots are not applicable, provides verification steps, identifies the change as a chore,… | | Docstring Coverage | | No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check. Docstring coverage is scoped to functions touched by this diff. Analyzed 0 functions across 0… | | Linked Issues check | | Check skipped because no linked issues were found for this pull request. | | Out of Scope Changes check | | Check skipped because no linked issues were found for this pull request. | Full details: Description check **Explanation** The description is complete and relevant. It explains the context, lists the removed sources, states that screenshots are not applicable, provides verification steps, identifies the change as a chore, and includes the required checklist items. Full details: Docstring Coverage **Explanation** No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check. Docstring coverage is scoped to functions touched by this diff. Analyzed 0 functions across 0 files. (2 skipped: 2 unsupported.) ✨ Finishing Touches 📝 Generate docstrings 🧪 Generate unit tests (beta) Comment `@the reviewer help` to get the list of available commands.
#8630
2026-10-20 19:31
coderabbit
inline · P2
frontend/src/components/V3_MIGRATION.md:62docs
**Reconcile the `NoticeBannerV2` consumer count.** This row reports `2 / 0` consumers, but the deletion proof says no incoming implementation edge or external symbol reference remained. `COMPONENT_LIFECYCLE.md` also records no consumers at the same commit. Set the count to `0 / 0`, or identify and explain the two historical files. Otherwise, the ledger provides contradictory deletion evidence.
#8630
2026-10-20 19:31
coderabbit
review-body
—summary
**Actionable comments posted: 1**
#8630
2026-10-20 19:32
cubic
review-body
—summary
**No issues found** across 7 files View guided diff | Turn on auto-fix | Re-trigger the reviewer
#8630
2026-10-20 19:37
greptile
issue-comment
—summary
Confidence Score: 5/5 **[Low impact]** No blocking issues were identified. Executed checks support that the deleted modules have no surviving consumers and that removing the empty family export leaves the public V2 exports unchanged. Summary Removes unused UI code without changing consumers: - The V2 catalog no longer exposes the unused NoticeBannerV2 banner. - The admin area drops its unused V3 adapter files. No actionable issues were found. Before-and-after reference and module-resolution checks found no surviving consumers of the deleted modules and confirmed that the public V2 barrel exports remain unchanged. Vite transformation and import-resolution checks passed for the barrel and
… the rest (699 characters)production dynamic-import callers and targets. The full frontend typecheck passed. The prior whitespace check also passed; a production build and browser tests were not run. Acknowledged by PierreOucif: the global `--assert-zero` gate still exits 1 because live legacy families remain. This is an intentional partial migration, not completion of the migration or establishment of a new workflow acceptance gate. No blocking issues were identified. The checked removals are safe to merge from a code-review perspective; merge authorization remains pending as stated in the PR description. Reviews (1) · Last reviewed commit: "chore(ui): delete orphan notice family a..." · Reviewed by the reviewer
#8646
2026-10-20 19:38
coderabbit
issue-comment
—summary
No actionable comments were generated in the recent review. 🎉 📝 Walkthrough The signing and verification modals now use v3 Dialog components. Each modal owns the operation state and passes it to its form. Dirty unsent input triggers a discard confirmation when a close is requested. The forms use DialogBody and DialogFooter. **Changes: KMS signing and verification dialogs** |Layer / File(s)|Summary| |:---|:---| |**Signing form and discard flow** `frontend/src/pages/kms/OverviewPage/components/CmekSignModal.tsx`|`CmekSignModal` owns the signing state and passes it to `SignForm`. The form resets signing results and reports dirty unsent input. Dirty close requests open a discard confirmati
… the rest (4654 characters)on; clean or completed forms close directly. The form and footer submit action use the v3 dialog layout.| |**Verification form and discard flow** `frontend/src/pages/kms/OverviewPage/components/CmekVerifyModal.tsx`|`CmekVerifyModal` owns the verification state and passes it to `VerifyForm`. The form resets verification results and reports dirty unsent input. Dirty close requests open a discard confirmation. Verification results and input fields remain conditionally displayed in the v3 dialog layout.| **Priority:** ⬇️ Low **Estimated code review effort:** 3 (Moderate) | ~20 minutes **Change:** Feature ### Sequence Diagram(s) ```mermaid sequenceDiagram participant SignForm participant CmekSignModal participant useDiscardChangesGuard participant DiscardChangesAlertDialog SignForm->>CmekSignModal: Report dirty unsent input CmekSignModal->>useDiscardChangesGuard: Check close request useDiscardChangesGuard->>DiscardChangesAlertDialog: Open discard confirmation DiscardChangesAlertDialog->>CmekSignModal: Confirm discard CmekSignModal->>CmekSignModal: Close dialog ``` ```mermaid sequenceDiagram participant VerifyForm participant CmekVerifyModal participant useDiscardChangesGuard participant DiscardChangesAlertDialog VerifyForm->>CmekVerifyModal: Report dirty unsent input CmekVerifyModal->>useDiscardChangesGuard: Check close request useDiscardChangesGuard->>DiscardChangesAlertDialog: Open discard confirmation DiscardChangesAlertDialog->>CmekVerifyModal: Confirm discard CmekVerifyModal->>CmekVerifyModal: Close dialog ``` **Suggested reviewers:** `bernie-g` Merge Risk: ⚪ Minimal · up to 04358 The dialog migration has no established merge-blocking issue after normal checks. The identified text and display issues can be addressed separately. Security architecture risk: 🔵 Low · up to 04358 The change is confined to signing and verification dialogs. Request formats and server-side permission checks remain unchanged, and the new controls protect pending operations and unsent input. No introduced security issue was established, but interruption and rapid-reopen behavior were not fully verified. **Retained concerns** No architecture-level concerns identified. Security review details **Security Blast Radius** - _inferred_ — Each request targets one selected key, while effective authority remains the caller's existing project-level Sign or Verify permission over customer-managed keys. Moving browser mutation ownership does not itself grant additional tenant, project, or key authority. **Trust Boundaries and Controls** - _observed_ — Browser-controlled key IDs, data, signatures, and algorithms cross an authenticated API boundary. The examined routes require authentication, UUID key IDs, enumerated algorithms, Base64 validation, payload limits, and rate limiting. Before KMS invocation, the service rejects missing, reserved, non-customer-managed, or disabled keys and requires the corresponding permission in the key's project. These controls predate this PR. Pre-merge checks: 5 | Check name | Status | Explanation | | :------------------------ | :--------------------------------------------------------------------------------------------: | :-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | Title check | | The title clearly and concisely describes the primary change: migrating KMS signing dialogs to V3 dialog shells. It also follows the conventional commit format. | | Description check | | The description includes all required sections, explains the context and scope, provides screenshots, verification steps, change type, checklist status, and testing details. | | Docstring Coverage | | No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check. Docstring coverage is scoped to functions touched by this diff. Analyzed 0 functions across 2… | | Linked Issues check | | Check skipped because no linked issues were found for this pull request. | | Out of Scope Changes check | | Check skipped because no linked issues were found for this pull request. | Full details: Docstring Coverage **Explanation** No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check. Docstring coverage is scoped to functions touched by this diff. Analyzed 0 functions across 2 files. ✨ Finishing Touches 📝 Generate docstrings 🧪 Generate unit tests (beta) Comment `@the reviewer help` to get the list of available commands.
#8646
2026-10-20 19:46
cubic
review-body
—summary
**No issues found** across 2 files View guided diff | Turn on auto-fix | Re-trigger the reviewer
#8646
2026-10-20 19:52
greptile
issue-comment
—summary
Confidence Score: 5/5 **[Medium impact]** No blocking defect was identified; the available evidence supports merging, with the incomplete algorithm-popup interaction check noted above. Summary Moves the KMS Sign and Verify dialogs to the V3 header, body, and footer composition. - The Sign Data dialog uses the V3 shell and guards unsent input. - The Verify Signature dialog uses the V3 shell and guards unsent input. No actionable issues or merge-blocking defects were identified in the reviewed changes. Acknowledged by PierreOucif: inherited Controller ref/onBlur omissions and the `decodeBase64(data).toString()` preview remain unchanged. Temporary toast overlap with a short footer is also
… the rest (552 characters) documented by PierreOucif as known; normal pointer departure restores its timeout. ### the check validation blocked Independent algorithm-popup interaction checks did not complete. The disposable browser harness encountered TSX transformation and source-import interception problems, and completed results were not collected after correction. Popup selection, Escape isolation, and trigger-focus restoration remain independently unverified. Reviews (1) · Last reviewed commit: "improvement(ui): compose KMS signing dia..." · Reviewed by the reviewer
#8647
2026-10-20 19:55
coderabbit
issue-comment
—summary
📝 Walkthrough Certificate Manager enrollment and collection dialogs now use v3 platform components. The changes update form validation, accessibility, and pending-state handling, and replace deletion modals with confirmation dialogs. **Changes: Certificate Manager UI** |Layer / File(s)|Summary| |:---|:---| |**Enrollment form and sheet** `frontend/src/pages/cert-manager/CertificatesPage/components/CertificateTemplateEnrollmentModal.tsx`|The enrollment form now uses v3 sheet components, generated control IDs, inline validation, and pending-aware actions. Reset behavior responds to popup state. The CA-chain controls are hidden and cleared when bootstrap certificate validation is disabled.|
… the rest (4940 characters)|**Add collection item** `frontend/src/pages/cert-manager/PkiCollectionDetailsByIDPage/components/AddPkiCollectionItemModal.tsx`|The add-item dialog now uses v3 form components and requires an item selection. Certificate loading uses the project ID, and the form displays validation and pending states.| |**Deletion dialogs and collection actions** `frontend/src/pages/cert-manager/DiscoveryPage/components/DeleteDiscoveryModal.tsx`, `frontend/src/pages/cert-manager/DiscoveryPage/components/DeleteInstallationModal.tsx`, `frontend/src/pages/cert-manager/PkiCollectionDetailsByIDPage/components/PkiCollectionItemsSection.tsx`|The deletion flows now use `DeleteConfirmDialog`. The collection section also uses the v3 icon button and adds an irreversible-action description.| **Priority:** ⬇️ Low **Estimated code review effort:** 3 (Moderate) | ~20 minutes **Change:** Refactor **Suggested reviewers:** `carlosmonastyrski`, `andrewhuhh` Merge Risk: 🔵 Low · up to 034d6 Assistive-technology users may not be able to identify which selection error belongs to the focused control. This is a bounded accessibility issue that should be fixed, but it does not block the collection workflow. Security architecture risk: 🔵 Low · up to 034d6 The changes remain confined to existing Certificate Manager workflows and preserve the inspected permission checks, mutation targets, secret masking, and pending-action protections. No introduced security weakness was demonstrated. Navigation during an active request and independent server-side enforcement remain unverified. **Retained concerns** No architecture-level concerns identified. Security review details **Security Blast Radius** - _inferred_ — The inspected frontend reach remains existing certificate-template configuration, project-scoped item discovery, collection membership, and selected discovery or installation records. Resource identities still come from parent selection or current-project context; no new credential, cross-service authority, or broader target selector was identified in these callers. This does not establish backend authorization enforcement. **Trust Boundaries and Controls** - _observed_ — The enrollment entry action retains its parent-owned template Edit permission and EST entitlement checks. Discovery deletion, installation deletion, and collection-item removal retain their respective permission-gated entry actions. Typed confirmation and required selection are interaction safeguards, not substitutes for server authorization. **Resilience and Maintainability Implications** - _observed_ — The new deletion contract requires an exact nonempty confirmation key, tracks the returned promise, disables confirmation and Cancel while pending, and refuses pending dismissal. Form callers pass submitting state into shared overlays and buttons, which disable close controls and submit buttons and guard Escape and outside dismissal. Delete pending state clears on both success and failure, while parent callbacks retain ownership of successful closure. Pre-merge checks: 5 | Check name | Status | Explanation | | :------------------------ | :--------------------------------------------------------------------------------------------: | :-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | Title check | | The title clearly identifies the main change: migrating cert-manager PKI overlays to V3. It is concise and follows the conventional commit format. | | Description check | | The description includes all required sections, explains the scope and behavior changes, provides screenshots, verification steps, change type, and a completed checklist. It also documents known verif… | | Docstring Coverage | | No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check. Docstring coverage is scoped to functions touched by this diff. Analyzed 0 functions across 5… | | Linked Issues check | | Check skipped because no linked issues were found for this pull request. | | Out of Scope Changes check | | Check skipped because no linked issues were found for this pull request. | Full details: Description check **Explanation** The description includes all required sections, explains the scope and behavior changes, provides screenshots, verification steps, change type, and a completed checklist. It also documents known verification limits. Full details: Docstring Coverage **Explanation** No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check. Docstring coverage is scoped to functions touched by this diff. Analyzed 0 functions across 5 files. ✨ Finishing Touches 📝 Generate docstrings 🧪 Generate unit tests (beta) Comment `@the reviewer help` to get the list of available commands.
#8647
2026-10-20 20:01
coderabbit
inline · P2
frontend/src/pages/cert-manager/PkiCollectionDetailsByIDPage/components/AddPkiCollectionItemModal.tsx:135defect-functional
**🎯 Functional Correctness** | **🟡 Minor** | **⚡ Quick win** **Associate each validation error with its select trigger.** `SelectTrigger` already exposes `aria-invalid` through `isError`. However, `Field` and `FieldError` do not create an error ID or apply `aria-describedby`. A user of assistive technology can receive the alert, but the focused select has no programmatic association with its specific error. Add an error ID and reference it from both triggers. Suggested fix ```diff <SelectTrigger id={itemInputId} ref={field.ref} onBlur={field.onBlur} className="w-full"
… the rest (359 characters) isError={Boolean(error)} + aria-describedby={error ? `${itemInputId}-error` : undefined} > ... - <FieldError>{error?.message}</FieldError> + <FieldError id={`${itemInputId}-error`}>{error?.message}</FieldError> ``` Apply the same change to the certificate selector.
#8647
2026-10-20 20:01
coderabbit
review-body
—summary
**Actionable comments posted: 1**
#8647
2026-10-20 20:03
cubic
review-body
—summary
**No issues found** across 5 files View guided diff | Turn on auto-fix | Re-trigger the reviewer
#8647
2026-10-20 20:10
greptile
issue-comment
—summary
Confidence Score: 5/5 **[Medium impact]** No merge-blocking issues were identified within the reviewed scope; focused browser checks support the enrollment and collection flows, subject to the stated limits. Summary Migrates five PKI consumers to supported V3 controls and overlays. - EST enrollment uses a scrollable Sheet with Save and Cancel in its footer. - Collection users add certificate authorities and certificates through a V3 dialog. - Collection item removal uses a typed confirmation dialog. - Discovery and installation deletion use the shared typed-confirmation dialog. No actionable issues were found. Isolated browser checks using the changed components, real form handling and s
… the rest (1046 characters)hared controls passed 19 EST checks covering create/update payloads, passphrase validation and masking, failed-save retention, pending dismissal protection, reset behavior and short-height scrolling. Collection checks passed for both CA and certificate attachment, including required selection, selected-value retention on the first Escape, expected payloads, pending dismissal protection and close/reopen reset. The subsequent parent-closing Escape was checked after popup teardown; back-to-back Escape behavior during teardown remains unverified. These browser checks used fake domain hooks and persistence, not authenticated backend requests. The three removal prompts were not exercised in these checks. Acknowledgments from `PierreOucif`: the backend’s empty-ID error remains outside this change, and authenticated provider/permission checks remain unverified. No merge-blocking issues were identified within the reviewed scope. Reviews (1) · Last reviewed commit: "improvement(cert-manager): migrate clear..." · Reviewed by the reviewer
#8633
2026-10-20 20:12
coderabbit
issue-comment
—summary
📝 Walkthrough The share form now uses SecretInput for secret values, limits optional secret names to 50 characters, and displays a count of active sharing constraints. SecretInput supports multiline sizing and alignment. **Changes: Share secret form** |Layer / File(s)|Summary| |:---|:---| |**Multiline secret input** `frontend/src/components/v3/platform/SecretInput/SecretInput.tsx`|SecretInput accepts `isMultiline`. When enabled for the default variant, it applies a taller minimum height, top alignment, and vertical padding.| |**Share form behavior** `frontend/src/pages/public/ShareSecretPage/components/ShareSecretForm.tsx`|The form uses SecretInput for the secret value. Supplied values a
… the rest (2726 characters)re read-only and show a locked-value description. The optional name is limited to 50 characters. The options trigger displays “No constraints” or the count of active constraints.| **Priority:** ⬇️ Low **Estimated code review effort:** 2 (Simple) | ~10 minutes **Change:** Feature **Suggested reviewers:** `adilsitos` Merge Risk: 🔵 Low · up to e3e9c The share form remains usable through other focus paths, but clicking the lower part of the multiline field does not focus it. Pre-merge checks: 5 | Check name | Status | Explanation | | :------------------------ | :--------------------------------------------------------------------------------------------: | :-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | Title check | | The title uses conventional commit format and clearly summarizes the main Share a Secret improvements: masked values, name prefilling, and constraint counts. | | Description check | | The description includes context, UI screenshots reference, verification steps, change type, and checklist. It explains the masked value field, read-only stored values, constraint count, and editable … | | Docstring Coverage | | No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check. Docstring coverage is scoped to functions touched by this diff. Analyzed 0 functions across 2… | | Linked Issues check | | Check skipped because no linked issues were found for this pull request. | | Out of Scope Changes check | | Check skipped because no linked issues were found for this pull request. | Full details: Description check **Explanation** The description includes context, UI screenshots reference, verification steps, change type, and checklist. It explains the masked value field, read-only stored values, constraint count, and editable organization-level sharing behavior. Full details: Docstring Coverage **Explanation** No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check. Docstring coverage is scoped to functions touched by this diff. Analyzed 0 functions across 2 files. ✨ Finishing Touches 📝 Generate docstrings 🧪 Generate unit tests (beta) ### Usage-based review receipt - Mode: Continue automatically - Reviewed files: 2 - Waived: $0.50 (charged $0.00) - View usage details > [!NOTE] > This review exceeded your plan’s limits and used usage-based reviews—free during trial, billed after paid activation unless disabled. **Manage usage-based reviews**. Comment `@the reviewer help` to get the list of available commands.
#8633
2026-10-20 20:14
cubic
inline · P2
frontend/src/pages/public/ShareSecretPage/components/ShareSecretForm.tsx:228defect-functional
The header stops counting specific recipients when external email delivery is enabled, even though those recipients still receive separate links. Count any non-empty recipient list so the constraint total reflects the selected recipients. ```suggestion !isPublic && parseRecipients(watch("emails")).length > 0 ```
#8633
2026-10-20 20:14
cubic
review-body
—summary
**2 issues found** across 2 files Reply with feedback, questions, or to request a fix.View guided diff | Turn on auto-fix | Re-trigger the reviewer
#8633
2026-10-20 20:14
cubic
inline · P3
frontend/src/components/v3/platform/SecretInput/SecretInput.tsx:249defect-functional
The multiline shell is taller than its content, but `items-start` leaves the absolutely positioned textarea covering only the one-line child. Make the child stretch so clicks anywhere in the visible field focus it. ```suggestion variant === "default" && isMultiline && "min-h-16 items-stretch py-1", ```
#8633
2026-10-20 20:15
coderabbit
inline · P2
frontend/src/components/v3/platform/SecretInput/SecretInput.tsx:249defect-functional
**🎯 Functional Correctness** | **🟡 Minor** | **⚡ Quick win** **Make the textarea fill the multiline field.** When the value or placeholder occupies one line, `min-h-16` enlarges the outer container but not the inner wrapper. The absolutely positioned textarea covers only the wrapper. Clicking the lower part of the visible field therefore does not focus it. Give the inner wrapper a matching minimum height while preserving its ability to grow with multiline content.
#8633
2026-10-20 20:15
coderabbit
review-body
—summary
**Actionable comments posted: 1**
#8633
2026-10-20 20:31
greptile
issue-comment
—summary
Confidence Score: 5/5 **[Medium impact]** No merge-blocking issues were found in the reviewed changes. Summary The share form now masks secret values and lets users focus stored values without editing them. - The share form masks secret values and explains when a stored value is locked. - More Options shows how many sharing constraints are active. - Share names have a 50-character limit. Browser checks confirmed that multiline values reveal on focus, conceal on blur, and retain their exact content through submission. Stored values reveal through label or keyboard focus and conceal again on blur while remaining read-only. No actionable issues were found. The reviewed changes are safe to
… the rest (116 characters)merge. Reviews (1) · Last reviewed commit: "improvement(secret-sharing): masked shar..." · Reviewed by the reviewer
#8656
2026-10-20 20:32
coderabbit
issue-comment
—summary
📝 Walkthrough The unused V2 `SecretPathInput` implementation and barrel export were removed. The lifecycle and migration records now mark the V2 component as candidate removal. The independent V3 implementation and consumers remain unchanged; parity validation and merged-main removal are not certified. **Changes: V2 SecretPathInput removal** |Layer / File(s)|Summary| |:---|:---| |**Remove V2 component and record candidate status** `frontend/src/components/v2/SecretPathInput/SecretPathInput.tsx`, `frontend/src/components/v2/SecretPathInput/index.tsx`, `frontend/src/components/V3_MIGRATION.md`, `frontend/src/components/COMPONENT_LIFECYCLE.md`|The V2 implementation and barrel export are del
… the rest (3589 characters)eted. The migration record documents the reference audit and candidate removal. The lifecycle record distinguishes `SecretPathInput` removal from `SecretInput`, which remains blocked.| **Priority:** ⬇️ Low **Estimated code review effort:** 2 (Simple) | ~10 minutes **Change:** Other **Suggested reviewers:** `sheensantoscapadngan`, `andrewhuhh` Merge Risk: 🔵 Low · up to c1629 The change is otherwise low risk, but the candidate inventory should be corrected before merge so audit records remain reliable. Architecture risk: 🔵 Low · up to c1629 The change affects 1 system. **Changed systems:** `frontend` **Architecture concerns** No architecture-level concerns identified. Review details **Systems and components** - _observed_ — frontend (ui) was modified; 4 changed files map to changed impact. **Before / after behavior** - _observed_ — Modified behavior in frontend/src/components/COMPONENT_LIFECYCLE.md: `SecretInput` remains blocked, while `SecretPathInput` replaces the previous blocked status with a candidate-removal record describing its deleted unused V2 implementation and family index, reference audit, and unchanged independent V3 implementation. - _observed_ — Modified behavior in frontend/src/components/V3_MIGRATION.md: Updated the `SecretPathInput` migration status from an active V3 replacement requiring parity checks to deletion of the unused V2 implementation and family index, with no incoming references or root, dynamic, glob, CSS, or initializer ingress identified. - _observed_ — Modified behavior in frontend/src/components/v2/SecretPathInput/SecretPathInput.tsx: The `SecretPathInput` implementation was deleted, including its public wrapper, project-context integration, folder-query and suggestion state management, keyboard navigation, and popover-based folder suggestions. - _observed_ — Modified behavior in frontend/src/components/v2/SecretPathInput/index.tsx: The file’s `SecretPathInput` re-export was removed. Pre-merge checks: 5 | Check name | Status | Explanation | | :------------------------ | :--------------------------------------------------------------------------------------------: | :-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | Title check | | The title clearly and concisely describes the removal of the unused V2 SecretPathInput source and follows the Conventional Commits format. | | Description check | | The description is complete and relevant. It includes context, verification steps, change type, checklist status, documentation impact, and the scope of unchanged V3 consumers. | | Docstring Coverage | | No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check. Docstring coverage is scoped to functions touched by this diff. Analyzed 0 functions across 0… | | Linked Issues check | | Check skipped because no linked issues were found for this pull request. | | Out of Scope Changes check | | Check skipped because no linked issues were found for this pull request. | Full details: Docstring Coverage **Explanation** No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check. Docstring coverage is scoped to functions touched by this diff. Analyzed 0 functions across 0 files. (2 skipped: 2 unsupported.) ✨ Finishing Touches 📝 Generate docstrings 🧪 Generate unit tests (beta) Comment `@the reviewer help` to get the list of available commands.
#8656
2026-10-20 20:36
coderabbit
inline · P2
frontend/src/components/V3_MIGRATION.md:70docs
**🗄️ Data Integrity & Integration** | **🟡 Minor** | **⚡ Quick win** **Synchronize the candidate inventory totals.** Line 83 presents the totals as current candidate-worktree numbers. The candidate deletes one family and two V2 files, so update the totals to 34 families and 70 files. Suggested fix ```diff -First wave deletes **17 families / 35 files** plus their root exports. The candidate retains **35 families / 72 V2 files including the barrel**, and the same 312 external consumer files. These are candidate-worktree numbers until the PR is explicitly merged; the baseline does not change because a PR was opened. +First wave deletes **17 families / 35 files** plus their root exports. Th
… the rest (243 characters)e candidate retains **34 families / 70 V2 files including the barrel**, and the same 312 external consumer files. These are candidate-worktree numbers until the PR is explicitly merged; the baseline does not change because a PR was opened. ```
#8656
2026-10-20 20:36
coderabbit
review-body
—summary
**Actionable comments posted: 1**
#8656
2026-10-20 20:38
cubic
inline · P3
frontend/src/components/V3_MIGRATION.md:70docs
Update the candidate inventory totals to 34 families and 70 files; this retirement removes one family and two V2 files, making the current totals stale.
#8656
2026-10-20 20:38
cubic
review-body
—summary
**1 issue found** across 4 files Reply with feedback, questions, or to request a fix.View guided diff | Turn on auto-fix | Re-trigger the reviewer
#8656
2026-10-20 20:41
greptile
issue-comment
—summary
Confidence Score: 5/5 **[Low impact]** Safe to merge within the stated source-retirement scope; current consumers continue to resolve to the unchanged V3 implementation. Summary Deletes the unused V2 `SecretPathInput` implementation and family index. - The unused V2 secret-path input is removed from the candidate. No actionable issues found. Dependency audits and compiler-resolved reference checks passed on both base and head. Frontend type checks passed on both revisions with a 6 GiB Node heap limit. Lint was not run. The change is safe to merge within this source-retirement scope. Reviews (1) · Last reviewed commit: "chore(ui): remove unused V2 SecretPathIn..." · Reviewed by the revie
… the rest (3 characters)wer
#8655
2026-10-20 20:43
coderabbit
issue-comment
—summary
No actionable comments were generated in the recent review. 🎉 📝 Walkthrough Mfa.tsx now displays WebAuthn failed-attempt messages in a danger Alert with a close icon. The basic Error component has been removed. **Changes: WebAuthn failure alert** |Layer / File(s)|Summary| |:---|:---| |**Replace the failed-attempt display** `frontend/src/components/auth/Mfa.tsx`, `frontend/src/components/basic/Error.tsx`|Mfa.tsx replaces the Error display with a danger Alert and the existing remaining-attempt message. The basic Error component is removed.| **Priority:** ⬇️ Low **Estimated code review effort:** 2 (Simple) | ~5 minutes **Change:** Refactor **Suggested reviewers:** `andrewhuhh` Merge
… the rest (2303 characters)Risk: ⚪ Minimal · up to 7805b The new alert presentation has no identified merge-blocking risk. Pre-merge checks: 5 | Check name | Status | Explanation | | :------------------------ | :--------------------------------------------------------------------------------------------: | :-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | Description check | | The description includes the required Context, Screenshots, Steps to verify the change, Type, and Checklist sections. It documents the scope, verification results, limitations, and checklist selection… | | Title check | | The title uses the conventional commit format and clearly describes the main change: replacing the passkey failure presentation with the V3 alert. | | Docstring Coverage | | No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check. Docstring coverage is scoped to functions touched by this diff. Analyzed 0 functions across 1… | | Linked Issues check | | Check skipped because no linked issues were found for this pull request. | | Out of Scope Changes check | | Check skipped because no linked issues were found for this pull request. | Full details: Description check **Explanation** The description includes the required Context, Screenshots, Steps to verify the change, Type, and Checklist sections. It documents the scope, verification results, limitations, and checklist selections clearly. Full details: Docstring Coverage **Explanation** No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check. Docstring coverage is scoped to functions touched by this diff. Analyzed 0 functions across 1 files. ✨ Finishing Touches 📝 Generate docstrings 🧪 Generate unit tests (beta) ### Usage-based review receipt - Mode: Continue automatically - Reviewed files: 1 - Waived: $0.25 (charged $0.00) - View usage details > [!NOTE] > This review exceeded your plan’s limits and used usage-based reviews—free during trial, billed after paid activation unless disabled. **Manage usage-based reviews**. Comment `@the reviewer help` to get the list of available commands.
#8655
2026-10-20 20:44
cubic
review-body
—summary
**No issues found** across 2 files View guided diff | Turn on auto-fix | Re-trigger the reviewer
#8655
2026-10-20 20:56
greptile
issue-comment
—summary
Confidence Score: 4/5 **[Low impact]** The change reaches only the passkey MFA failure display\. Safe to merge with one non-blocking readability concern in the passkey failure warning. Findings 1. **Restore error text contrast** Summary Replaces the passkey failure message in `Mfa.tsx` with the V3 danger Alert and deletes the unused Error helper. - Passkey failures now appear in the V3 danger alert. ### the check validation blocked Evidence upload was unavailable because the environment exposed no artifact-upload tool or upload endpoint. Browser execution completed successfully, and scripts, output, and screenshots were preserved locally. Authenticated MFA and passkey-provider accept
… the rest (131 characters)ance were not tested. Reviews (1) · Last reviewed commit: "improvement(auth): use V3 alert for pass..." · Reviewed by the reviewer
#8655
2026-10-20 20:56
greptile
inline · P2
frontend/src/components/auth/Mfa.tsx:433defect-functional
**Restore error text contrast** The new `AlertDescription` inherits the danger variant's 80%-opacity text color. Rendering the exact replacement with the application styles reduces the remaining-attempts message's contrast from 5.09:1 to 3.54:1 in dark mode, below the 4.5:1 minimum for normal-sized text. The additional light-theme check also falls from 5.93:1 to 4.12:1. This is a non-blocking readability regression for users reading the passkey failure warning, particularly those with low vision. Use a sufficiently contrasting description color while retaining the V3 alert. Note: If this suggestion doesn't match your team's coding style, reply to this and let me know. I'll remember it for
… the rest (46 characters)next time! Ran code and verified through TREX
#8661
2026-10-20 20:58
coderabbit
issue-comment
—summary
No actionable comments were generated in the recent review. 🎉 📝 Walkthrough The changes make the action bar’s clear control icon-only by default, retain the non-icon-only control in the secret dashboard commit form, and add conditional right padding to toasts with a close button. **Changes: Selected Action Bar Clear Control** |Layer / File(s)|Summary| |:---|:---| |**Clear control default and override** `frontend/src/components/v3/generic/SelectedActionBar/SelectedActionBar.tsx`, `frontend/src/pages/secret-manager/SecretDashboardPage/components/CommitForm/CommitForm.tsx`|`SelectedActionBar` defaults `iconOnlyClear` to `true`. The secret dashboard commit form sets it to `false`.| **Toas
… the rest (2654 characters)t Close Button Padding** |Layer / File(s)|Summary| |:---|:---| |**Conditional toast padding** `frontend/src/components/v3/generic/Toast/Toast.tsx`|Toast styling adds right padding when a close button descendant is present.| **Priority:** ⬇️ Low **Estimated code review effort:** 2 (Simple) | ~8 minutes **Change:** Bug fix **Suggested reviewers:** `andrewhuhh` Merge Risk: ⚪ Minimal · up to 33811 The shared UI changes are ready to merge with no established blocking risk. Pre-merge checks: 4 · 1 Failed checks: 1 warning | Check name | Status | Explanation | Resolution | | :---------------- | :----------------------------------------------------------------------------------------------: | :---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | :--------------------------------------------------------------------------------- | | Docstring Coverage | | Docstring coverage is 0.00% which is insufficient. The required threshold is 80.00%. Docstring coverage is scoped to functions touched by this diff. Analyzed 1 functions across 3 files. | Write docstrings for the functions missing them to satisfy the coverage threshold. | Passed checks: 4 passed | Check name | Status | Explanation | | :------------------------ | :--------------------------------------------------------------------------------------------: | :-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | Title check | | The title clearly summarizes both main UI fixes and follows the conventional commit format. | | Description check | | The description is complete and matches the required template. It explains the problems and solutions, lists affected and unchanged surfaces, includes screenshots, verification steps, selected change … | | Linked Issues check | | Check skipped because no linked issues were found for this pull request. | | Out of Scope Changes check | | Check skipped because no linked issues were found for this pull request. | Full details: Description check **Explanation** The description is complete and matches the required template. It explains the problems and solutions, lists affected and unchanged surfaces, includes screenshots, verification steps, selected change types, and checklist results. ✨ Finishing Touches: 💡 1 📝 Generate docstrings: 💡 🧪 Generate unit tests (beta) Comment `@the reviewer help` to get the list of available commands.
#8661
2026-10-20 20:59
cubic
review-body
—summary
**No issues found** across 3 files View guided diff | Turn on auto-fix | Re-trigger the reviewer
#8661
2026-10-20 21:10
greptile
issue-comment
—summary
Confidence Score: 5/5 **[Medium impact]** Safe to merge. Browser checks confirmed the intended toast spacing and selection-control behavior without identifying a regression. Summary Updates two shared UI defaults: - Toasts leave room for the close button when it appears. - Bulk-selection bars show an X clear button by default. Browser checks confirmed that long toast titles remain clear of the close button at desktop and mobile widths. Loading toasts and toasts with closeButton: false retain their original padding and layout. The selection-clear button has the “Unselect All” accessible name and tooltip, and clearing works by click, Escape, and keyboard activation. The explicit “Discard”
… the rest (208 characters)text override remains functional. No actionable issues were found. The changes are safe to merge. Reviews (1) · Last reviewed commit: "fix(ui): reserve toast close-button spac..." · Reviewed by the reviewer
#8659
2026-10-20 21:11
coderabbit
issue-comment
—summary
No actionable comments were generated in the recent review. 🎉 📝 Walkthrough SurfaceSelect now uses Combobox instead of FilterableSelect. It retains multi-select behavior, maps changes from the selected options, and sets a search accessibility label. **Changes: Surface selector** |Layer / File(s)|Summary| |:---|:---| |**Combobox integration** `frontend/src/components/proxied-services/forms/SurfaceSelect.tsx`|SurfaceSelect replaces FilterableSelect with Combobox. It maps option changes from the selected option array and sets the search accessibility label to "Select surfaces".| **Priority:** ⬇️ Low **Estimated code review effort:** 2 (Simple) | ~5 minutes **Change:** Refactor **Sugge
… the rest (2408 characters)sted reviewers:** `saifsmailbox98` Merge Risk: ⚪ Minimal · up to 7d8de The selector migration preserves the existing multi-select behavior and has no identified merge-blocking risk. Pre-merge checks: 5 | Check name | Status | Explanation | | :------------------------ | :--------------------------------------------------------------------------------------------: | :-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | Title check | | The title clearly and concisely describes the main change: migrating the proxied-services surface picker to Combobox. It follows the conventional commit format. | | Description check | | The description includes context, screenshots, verification steps, change type, checklist status, test results, and explicit test limitations. It aligns with the repository template and the pull reque… | | Docstring Coverage | | No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check. Docstring coverage is scoped to functions touched by this diff. Analyzed 0 functions across 1… | | Linked Issues check | | Check skipped because no linked issues were found for this pull request. | | Out of Scope Changes check | | Check skipped because no linked issues were found for this pull request. | Full details: Description check **Explanation** The description includes context, screenshots, verification steps, change type, checklist status, test results, and explicit test limitations. It aligns with the repository template and the pull request objectives. Full details: Docstring Coverage **Explanation** No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check. Docstring coverage is scoped to functions touched by this diff. Analyzed 0 functions across 1 files. ✨ Finishing Touches 📝 Generate docstrings 🧪 Generate unit tests (beta) ### Usage-based review receipt - Mode: Continue automatically - Reviewed files: 1 - Waived: $0.25 (charged $0.00) - View usage details > [!NOTE] > This review exceeded your plan’s limits and used usage-based reviews—free during trial, billed after paid activation unless disabled. **Manage usage-based reviews**. Comment `@the reviewer help` to get the list of available commands.
#8659
2026-10-20 21:12
cubic
review-body
—summary
**No issues found** across 1 file View guided diff | Turn on auto-fix | Re-trigger the reviewer
#8659
2026-10-20 21:25
greptile
issue-comment
—summary
Confidence Score: 5/5 **[Low impact]** Safe to merge within the reviewed scope; the actual picker preserved controlled form values and passed the exercised browser interactions. Summary Replaces `FilterableSelect` with `Combobox` in `SurfaceSelect`, preserving the four surface options, their order, controlled enum-array mapping, clear behavior, and disabled/error props. Adds the accessible search name “Select surfaces.” No actionable issues found. Browser checks using the actual components, production styles, and a React Hook Form controlled field passed pointer and keyboard selection, external edit values, chip removal, Backspace, clearing to an empty array, disabled/error states, focus
… the rest (601 characters) retention, and Escape behavior. The migrated picker passed all 20 main interaction checks; the previous picker passed all 18 applicable comparison checks. Supplemental edit-transition and computed error-style checks also passed. Nine shared Combobox tests and scoped lint passed. Safe to merge within this picker-migration scope. Authenticated create/edit wizard submission and provider acceptance were not exercised. - The surface picker now uses the V3 Combobox for multiple selections. Reviews (1) · Last reviewed commit: "improvement(proxied-services): migrate s..." · Reviewed by the reviewer
#8587
2026-10-20 21:27
coderabbit
issue-comment
—summary
No actionable comments were generated in the recent review. 🎉 📝 Walkthrough Cron enqueueing now tracks each job’s last scheduled fire in Redis. The enqueue logic handles existing run hashes and applies a threshold to first fires. New tests cover restart deduplication, missing timestamps, and first-fire boundaries. **Changes: Cron fire deduplication** |Layer / File(s)|Summary| |:---|:---| |**Persistent fire tracking and enqueue logic** `backend/src/lib/cron/cron-job.ts`|The enqueue script uses a persistent per-job last-fire key, handles existing run hashes, and applies a two-interval threshold to first fires. The caller passes three Redis keys and a migration eligibility flag.| |**Enque
… the rest (5475 characters)ue behavior validation** `backend/e2e-test/cron-job.spec.ts`, `backend/src/lib/cron/cron-job.test.ts`|End-to-end tests cover restart deduplication and first-fire cases. Unit tests identify three-key EVAL calls and parse the updated argument positions.| **Priority:** ➖ Normal **Estimated code review effort:** 3 (Moderate) | ~25 minutes **Change:** Bug fix **Suggested reviewers:** `victorvhs017` Merge Risk: ⚪ Minimal · up to 04362 The change makes cron fire deduplication persistent across restarts, with tests covering restart, backfill, and first-fire behavior. No merge-blocking risk was identified. Security architecture risk: 🟡 Moderate · up to 04362 Persistent deduplication prevents repeated scheduling after restarts, but it also prevents recovery of an unexecuted fire after its temporary run state expires. For weekly encryption-key cleanup, this can postpone security maintenance until the following week. Existing worker controls remain in place, and the inspected production wiring does not introduce a new externally accessible entrypoint. **Retained concerns** - **Medium · reliability · inferred:** The durable marker can suppress recovery of work that never executed. If a fire is enqueued, all relevant workers are interrupted until its run hash expires, and a worker restarts before the next scheduled fire, LAST_KEY causes enqueueing to return AlreadyEnqueued while processing removes the expired pending entry. The base implementation could recreate that latest fire after restart. KMS root-key cleanup runs weekly with only one hour of run-state retention, so this failure sequence can postpone retirement of eligible obsolete keys until the following week. Lease reclaim and retries do not recover a run whose hash has disappeared. Security review details **Security Blast Radius** - _inferred_ — The recovery concern follows a job's shared Redis name and prefix across workers, not an individual request or tenant. For KMS cleanup, its downstream effect is delayed instance-level obsolete-key retirement. The evidence does not establish credential disclosure, privilege escalation, or attacker control over this transition. **Security Findings and Attack Paths** - _inferred_ — The identified concern is an interruption-and-restart recovery regression affecting security maintenance, not a verified attacker-driven exploit. It requires run-state expiry before successful execution and restart before the next scheduled fire. **Trust Boundaries and Controls** - _observed_ — The inspected production path supplies Redis and locking authority during initialization. Run processing selects a handler from the in-process registration map and retains lease-based execution. The changed Lua script does not introduce request-supplied handler selection or new execution credentials. **Resilience and Maintainability Implications** - _observed_ — The added integration scenarios cover suppression after a completed run's hash disappears, legacy backfill, overdue first-fire skipping, and the just-due migration exception. The expired-hash restart scenario explicitly completes the first handler before deleting its run state, so it does not establish recovery of an unexecuted fire. **Hardening Proposals** - _proposed_ — Define an explicit recovery contract that distinguishes enqueue acknowledgement from terminal execution outcome, then retain sufficient durable state to reconcile an expired unclaimed run without blindly replaying completed work. Validate the interruption-and-restart sequence using the weekly cleanup retention window. Pre-merge checks: 5 | Check name | Status | Explanation | | :------------------------ | :--------------------------------------------------------------------------------------------: | :-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | Title check | | The title uses conventional commit format and clearly summarizes the cron deduplication and first-fire behavior changes. | | Description check | | The description provides detailed context, verification steps, change type, rollout behavior, known rollback risk, and test results. The required structure is substantially complete. | | Docstring Coverage | | No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check. Docstring coverage is scoped to functions touched by this diff. Analyzed 0 functions across 3… | | Linked Issues check | | Check skipped because no linked issues were found for this pull request. | | Out of Scope Changes check | | Check skipped because no linked issues were found for this pull request. | Full details: Docstring Coverage **Explanation** No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check. Docstring coverage is scoped to functions touched by this diff. Analyzed 0 functions across 3 files. ✨ Finishing Touches 📝 Generate docstrings 🧪 Generate unit tests (beta) ### Usage-based review receipt - Mode: Continue automatically - Reviewed files: 3 - Waived: $0.75 (charged $0.00) - View usage details > [!NOTE] > This review exceeded your plan’s limits and used usage-based reviews—free during trial, billed after paid activation unless disabled. **Manage usage-based reviews**. Comment `@the reviewer help` to get the list of available commands.
#8587
2026-10-20 21:34
cubic
inline · P2
backend/src/lib/cron/cron-job.ts:117defect-functional
This timestamp watermark assumes a job’s cron pattern never changes. Changing a job to a less-frequent pattern can leave its next occurrence earlier than the stored `last`, so this comparison suppresses it; include a pattern/version fingerprint or reset the watermark when the pattern changes.
#8587
2026-10-20 21:34
cubic
review-body
—summary
**2 issues found** across 3 files Reply with feedback, questions, or to request a fix.View guided diff | Turn on auto-fix | Re-trigger the reviewer
#8587
2026-10-20 21:34
cubic
inline · P2
backend/src/lib/cron/cron-job.ts:137defect-concurrency
This skip is not visible to old pods: their enqueue script only checks `runKey`. During a mixed-version rollout, an old slot holder can enqueue this fire after this branch, defeating first-fire suppression; write a tombstone run hash with a TTL or defer this behavior until old pods are gone.
#8587
2026-10-20 21:35
greptile
issue-comment
—summary
Confidence Score: 4/5 **[Medium impact]** The change reaches scheduled jobs across the backend’s general\-worker workflow\. The reported test coverage gap is non-blocking; no blocking production defect is established. Findings 1. **Next fire is not tested** Summary Adds a permanent per-job timestamp so expired run hashes no longer allow duplicate fires. - Cron jobs keep a lasting record of each fire they enqueue. - New cron jobs wait for their next scheduled fire. PierreOucif explicitly documents skipping old fires on first registration as intentional, so newly registered jobs do not immediately run an old fire on deploy. PierreOucif also documents allowing just-due fires during migra
… the rest (613 characters)tion as intentional, to avoid dropping a boundary fire during the initial rollout. PierreOucif acknowledges the rollback/redeploy duplicate-fire limitation as known: if a rollback lasts beyond run-hash expiry, redeployment can enqueue that fire again; the PR describes this as existing main behavior rather than a regression. The remaining concern is a non-blocking test coverage gap in backend/e2e-test/cron-job.spec.ts. It does not establish a production scheduling defect and does not block merging. Reviews (1) · Last reviewed commit: "chore(cron): label the first-deploy migr..." · Reviewed by the reviewer
#8587
2026-10-20 21:35
greptile
inline · P2
backend/e2e-test/cron-job.spec.ts:586tests
**Next fire is not tested** This test checks that the old fire is skipped, but never moves the clock forward to check that the next fire runs. The unchanged test also passes against a scheduler double that continually skips fires, so it does not protect the promise that a newly registered job eventually runs. This is a non-blocking coverage gap: extend the test past the next minute and assert exactly one handler call. Note: If this suggestion doesn't match your team's coding style, reply to this and let me know. I'll remember it for next time! Artifacts **Executed source for the next-fire coverage check** - Uploaded the authored script that executes the repository callback against two i
… the rest (653 characters)solated scheduler doubles; HTTP 200 confirms the source was uploaded. **Existing test accepts a continually-skipping scheduler** - Uploaded the captured original execution showing both scheduler doubles pass, demonstrating the missing next-fire coverage. **Extended test rejects a continually-skipping scheduler** - Uploaded the captured extended execution showing the control passes and the continually-skipping double fails, supporting the proposed assertion. **Tracked files remain unchanged** - Uploaded the captured Git diff checks and proof-file hashes, confirming validation left tracked files unchanged. Ran code and verified through TREX
#8644
2026-10-20 21:37
coderabbit
issue-comment
—summary
No actionable comments were generated in the recent review. 🎉 📝 Walkthrough The universal auth login path now converts the client-secret usage count and configured limit to numbers before checking whether the limit has been reached. The existing revocation and unauthorized-error behavior remains unchanged. **Changes: Universal auth usage limit** |Layer / File(s)|Summary| |:---|:---| |**Normalize usage limit comparison** `backend/src/services/identity-ua/identity-ua-service.ts`|The login path compares numeric conversions of the client-secret usage count and limit before enforcing the limit.| **Priority:** ➖ Normal **Estimated code review effort:** 2 (Simple) | ~10 minutes **Change:**
… the rest (4701 characters) Bug fix · **Severity of issue fixed:** Medium **Suggested reviewers:** `thiago-as` Merge Risk: ⚪ Minimal · up to 4e4ff Universal Auth login now compares client-secret usage counts and limits as numbers. Secrets with multi-digit usage limits will no longer be revoked early. Enforcement at the configured limit still revokes the secret and rejects the login. No merge-blocking risk was identified in this change. Security architecture risk: ⚪ Minimal · up to 4e4ff The fix restores numeric usage-limit enforcement without changing credential validation, token authority, or public interfaces. No material security risk introduced or materially worsened by this change was identified. **Retained concerns** No architecture-level concerns identified. Security review details **Security Blast Radius** - _inferred_ — The affected security outcome is token issuance using an existing machine-identity credential. Reaching the changed gate requires a matching secret and passing the existing IP and lockout checks. The conversion introduces no new identity, organization scope, or token privilege. **Security Findings and Attack Paths** - _inferred_ — Concurrent requests can pass the application-level snapshot check before their increments complete. The same opportunity exists at the base for a valid single-use secret, so correcting multi-digit comparisons does not introduce this transition gap or increase its maximum credential authority. Actual database-level prevention was not verified. **Trust Boundaries and Controls** - _observed_ — The public route applies its existing write rate limit and input schema. The service retains secret-hash verification, non-revoked lookup filtering, configured IP and lockout checks, expiration enforcement, and organization-membership checks before token issuance. **Resilience and Maintainability Implications** - _observed_ — For finite limits, login awaits the usage-update transaction before calling the token issuer. Login contains no compensation that refunds a committed use after later token-issuance or response failure. The exhaustion branch writes revocation before rejecting the request. These failure-ordering behaviors are unchanged. Pre-merge checks: 5 | Check name | Status | Explanation | | :------------------------ | :--------------------------------------------------------------------------------------------: | :-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | Title check | | The title clearly and concisely describes the main change and follows the conventional commit format. | | Description check | | The description explains the issue, technical cause, solution, verification steps, fix type, and checklist status. The omitted Screenshots section is not required because this change has no UI impact. | | Linked Issues check | | Issue `#8378` is closed and supplies historical context only, so it creates no active coding requirements. The reviewed diff nevertheless matches the issue context: the Universal Auth login check conver… | | Out of Scope Changes check | | The pull request changes only the usage-limit comparison in backend/src/services/identity-ua/identity-ua-service.ts. This change directly addresses the stated Universal Auth login bug and introduces n… | | Docstring Coverage | | No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check. Docstring coverage is scoped to functions touched by this diff. Analyzed 0 functions across 1… | Full details: Linked Issues check **Explanation** Issue `#8378` is closed and supplies historical context only, so it creates no active coding requirements. The reviewed diff nevertheless matches the issue context: the Universal Auth login check converts both usage values with Number() before it compares them, while it preserves the existing revocation behavior. Full details: Out of Scope Changes check **Explanation** The pull request changes only the usage-limit comparison in backend/src/services/identity-ua/identity-ua-service.ts. This change directly addresses the stated Universal Auth login bug and introduces no unrelated behavior or files. Full details: Docstring Coverage **Explanation** No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check. Docstring coverage is scoped to functions touched by this diff. Analyzed 0 functions across 1 files. ✨ Finishing Touches 📝 Generate docstrings 🧪 Generate unit tests (beta) Comment `@the reviewer help` to get the list of available commands.
#8644
2026-10-20 21:37
cubic
review-body
—summary
**No issues found** across 1 file View guided diff | Turn on auto-fix | Re-trigger the reviewer
#8644
2026-10-20 21:42
greptile
issue-comment
—summary
Confidence Score: 5/5 **[Medium impact]** Safe to merge within the reviewed scope: the changed comparison corrects premature revocation while preserving finite-limit enforcement and unlimited-secret behavior. Summary The PR converts client-secret usage counts and limits to numbers before comparing them. - Universal Auth login checks client-secret limits as numbers. Reviews (1) · Last reviewed commit: "fix(identity-ua): compare client secret ..." · Reviewed by the reviewer
#8667
2026-10-20 21:49
coderabbit
issue-comment
—summary
No actionable comments were generated in the recent review. 🎉 📝 Walkthrough The blackbox workflow now runs `make test-suites`. The Infisical readiness probe now has a two-minute startup timeout and retains its HTTP 200 matcher. **Changes: Blackbox test execution** |Layer / File(s)|Summary| |:---|:---| |**Select the test suites target** `.github/workflows/run-blackbox-tests.yml`|The `Run suites` step now invokes `make test-suites` instead of `make test-all`.| |**Set the readiness startup timeout** `tests/harness/infisical/ready.go`|`readyStrategy` now applies a two-minute startup timeout to the `/api/status` probe. The HTTP 200 requirement remains unchanged.| **Priority:** ⬇️ Low **Es
… the rest (2239 characters)timated code review effort:** 2 (Simple) | ~5 minutes **Change:** Bug fix **Suggested reviewers:** `akhilmhdh` Merge Risk: ⚪ Minimal · up to fa45e The black-box workflow now runs the intended suites, with a longer allowance for Infisical startup. No actionable merge risk remains after normal checks. Pre-merge checks: 5 | Check name | Status | Explanation | | :------------------------ | :--------------------------------------------------------------------------------------------: | :-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | Title check | | The title identifies the CI test failure fix and matches the main change. It contains spelling errors in "resovled" and "contraint," but remains clear and related. | | Description check | | The description explains the CI test failure, the test-suite change, and the timeout increase. It marks the change as a fix. The verification steps and several checklist items are incomplete, but the … | | Docstring Coverage | | Docstring coverage is 100.00% which is sufficient. The required threshold is 80.00%. Docstring coverage is scoped to functions touched by this diff. Analyzed 1 functions across 1 files. (1 skipped: 1 … | | Linked Issues check | | Check skipped because no linked issues were found for this pull request. | | Out of Scope Changes check | | Check skipped because no linked issues were found for this pull request. | Full details: Description check **Explanation** The description explains the CI test failure, the test-suite change, and the timeout increase. It marks the change as a fix. The verification steps and several checklist items are incomplete, but the description is mostly complete. Full details: Docstring Coverage **Explanation** Docstring coverage is 100.00% which is sufficient. The required threshold is 80.00%. Docstring coverage is scoped to functions touched by this diff. Analyzed 1 functions across 1 files. (1 skipped: 1 unsupported.) ✨ Finishing Touches 📝 Generate docstrings 🧪 Generate unit tests (beta) Comment `@the reviewer help` to get the list of available commands.
#8667
2026-10-20 21:50
cubic
inline · P2
tests/harness/infisical/ready.go:32defect-functional
This strategy now fails after two minutes even though the harness allows five minutes for the application’s 540 cold-start migrations. Keep the readiness timeout aligned with the five-minute startup allowance. ```suggestion WithStartupTimeout(5 * time.Minute) ```
#8667
2026-10-20 21:50
cubic
inline · P2
.github/workflows/run-blackbox-tests.yml:86tests
Switching to `make test-suites` (which runs only `./suites/...`) permanently removes the entire `harnesstest` package from CI. No other workflow runs `make test-harness`, `test-all`, or the harnesstest package (verified: the only match across `.github/workflows/` is this file), and `tests/Makefile` defines `test-suites: PKGS = ./suites/...` while `test-all` includes `./harnesstest/...`. This fixes the red build by deleting coverage rather than by the actual root-cause fix in this PR (the startup-timeout change in `tests/harness/infisical/ready.go`). The harness tests prove the test tooling itself and are the only CI check that catches harness regressions, so when the API is rewritten the har
… the rest (401 characters)ness can silently break without anything failing. Keep the harness covered by running `make test-suites && make test-harness` (each target starts/stops the stack with its own trap), or run the harnesstest packages as a separate step; at minimum document explicitly why `./harnesstest/...` is excluded so it doesn't look accidental. ```suggestion run: make test-suites && make test-harness ```
#8667
2026-10-20 21:50
cubic
review-body
—summary
**2 issues found** across 2 files Reply with feedback, questions, or to request a fix.View guided diff | Turn on auto-fix | Re-trigger the reviewer
#8667
2026-10-20 21:56
greptile
issue-comment
—summary
Confidence Score: 4/5 **[Low impact]** The changes stay within the blackbox test workflow and container startup wait\. Safe to merge with a non-blocking gap in harness regression coverage. Findings 1. **Harness tests stop running** Summary This PR separates the black-box suites from the harness tests and gives the application longer to start. - Blackbox CI runs the product suites without the harness tests. - The harness gives Infisical two minutes to become ready. The package-selection check confirmed that both harness regression packages stop running in CI, and the unit job does not replace that coverage. This is a non-blocking coverage gap; no blocking issue was established. Review
… the rest (102 characters)s (1) · Last reviewed commit: "fix: resovled test failing due to ci con..." · Reviewed by the reviewer
#8667
2026-10-20 21:56
greptile
inline · P2
.github/workflows/run-blackbox-tests.yml:86tests
**Harness tests stop running** Switching to `test-suites` removes the only CI run of `./harnesstest/...`. `test-unit` also excludes those packages, and no other workflow runs them. This leaves the isolated-instance and shared-harness regression checks without CI coverage, so regressions they catch can pass unnoticed. This is a non-blocking coverage concern: keep this job limited to suites, but run `test-harness` in a separate job so it does not compete for the same runner's resources. Note: If this suggestion doesn't match your team's coding style, reply to this and let me know. I'll remember it for next time! Artifacts **Evidence from the check** - Executed against both revisions to in
… the rest (593 characters)spect workflow invocations and probe actual Make package selection, demonstrating the harness coverage loss without launching Docker. **Command output from the check** - Captured the base workflow-selected Make recipe, complete workflow invocation inventory, and passing unit run; test-all selects both harness regression packages. **Command output from the check** - Captured the head workflow-selected Make recipe, complete workflow invocation inventory, and passing unit run; test-suites omits both harness packages with no replacement CI invocation. Ran code and verified through TREX
#8576
2026-10-20 21:58
coderabbit
issue-comment
—summary
No actionable comments were generated in the recent review. 🎉 📝 Walkthrough Secret folder queries now share search and tag predicates. The test harness adds project-role creation support, and integration tests cover tag-filtered secret listings, dashboard searches, and tag-based permissions. **Changes: Secret tag queries and permissions** |Layer / File(s)|Summary| |:---|:---| |**Shared search and tag predicates** `backend/src/services/secret-v2-bridge/secret-v2-bridge-dal.ts`|Added shared search and correlated tag-selection predicates. `countByFolderIds` and `findByFolderIds` use them for search and tag-slug filtering.| |**Role and secret test harness** `tests/clients/api/oapi-codegen.
… the rest (4976 characters)yaml`, `tests/clients/api/overlay.yaml`, `tests/docs/client.md`, `tests/docs/writing-tests.md`, `tests/fixture/role.go`, `tests/suites/secretmanager/secrets/helpers_test.go`|Added API generation configuration and an overlay, project-role fixture support, secret listing helpers, and related harness documentation.| |**Tag filtering and permission integration tests** `tests/suites/secretmanager/secrets/dashboard_test.go`, `tests/suites/secretmanager/secrets/permission_test.go`|Added coverage for dashboard search and listing with tag filters, allowed and denied tags, and conflicting allow and deny roles.| **Priority:** ➖ Normal **Estimated code review effort:** 3 (Moderate) | ~25 minutes **Change:** Bug fix **Suggested reviewers:** `akhilmhdh` Merge Risk: ⚪ Minimal · up to fdd91 No actionable issue was established in the changed secret-listing or test paths. The PR is mergeable after normal checks. Security architecture risk: 🔵 Low · up to fdd91 The change strengthens tag-based access checks while preserving the reviewed project, folder, and actor restrictions. No introduced security regression was established. Uncertainty remains about whether aggregate counts should reflect per-secret access restrictions. **Retained concerns** No architecture-level concerns identified. Security review details **Security Blast Radius** - _inferred_ — The affected exposure is authenticated secret listing within resolved project and folder scope, including keys, tags, metadata, and permission-governed values. The reviewed change preserves folder and shared-or-owner restrictions rather than granting access to another tenant or actor's personal secrets. **Security Findings and Attack Paths** - _inferred_ — The relevant attack path is choosing a benign tag or search term that removes a deny tag from the authorization input. Separating existence-based selection from tag hydration addresses that path in the reviewed queries. No introduced or materially worsened attack path was established. **Trust Boundaries and Controls** - _observed_ — Search and tag parameters refine candidate selection rather than confer authority. Service checks evaluate complete secret subjects through the existing permission helper, which supports both separate describe/read-value actions and the legacy combined action. The helper and service enforcement are unchanged between base and head. **Resilience and Maintainability Implications** - _observed_ — Aggregate counting remains separate from per-secret tag authorization. This predates the PR and is not retained as an introduced concern. Whether counts must conceal secrets denied by tag-specific policy remains an unresolved disclosure-policy question. Pre-merge checks: 4 · 1 Failed checks: 1 warning | Check name | Status | Explanation | Resolution | | :---------------- | :----------------------------------------------------------------------------------------------: | :-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | :--------------------------------------------------------------------------------- | | Docstring Coverage | | Docstring coverage is 0.00% which is insufficient. The required threshold is 80.00%. Docstring coverage is scoped to functions touched by this diff. Analyzed 7 functions across 5 files. (4 skipped: 4 … | Write docstrings for the functions missing them to satisfy the coverage threshold. | Passed checks: 4 passed | Check name | Status | Explanation | | :------------------------ | :--------------------------------------------------------------------------------------------: | :---------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | Title check | | The title uses the conventional commit format and clearly describes the main query change: returning all tags for access evaluation. | | Description check | | The description includes context, verification steps, change type, and checklist status. It also explains the tag-based permission fix and documents the test coverage. | | Linked Issues check | | Check skipped because no linked issues were found for this pull request. | | Out of Scope Changes check | | Check skipped because no linked issues were found for this pull request. | Full details: Docstring Coverage **Explanation** Docstring coverage is 0.00% which is insufficient. The required threshold is 80.00%. Docstring coverage is scoped to functions touched by this diff. Analyzed 7 functions across 5 files. (4 skipped: 4 unsupported.) **Resolution** Write docstrings for the functions missing them to satisfy the coverage threshold. ✨ Finishing Touches: 💡 1 📝 Generate docstrings: 💡 🧪 Generate unit tests (beta) Comment `@the reviewer help` to get the list of available commands.
#8576
2026-10-20 22:06
cubic
inline · P2
tests/fixture/role.go:26defect-functional
Calling `NewRole` without permissions encodes `permissions` as `null`, which the create-role API rejects because it expects an array. Encode an empty slice as `[]` for this case. ```suggestion body, err := json.Marshal(map[string]any{"slug": slug, "name": slug, "permissions": append([]Permission{}, permissions...)}) ```
#8576
2026-10-20 22:06
cubic
review-body
—summary
**1 issue found** across 10 files View guided diff | Turn on auto-fix | Re-trigger the reviewer
#8576
2026-10-20 22:06
greptile
issue-comment
—summary
Confidence Score: 5/5 **[Medium impact]** No blocking findings remain; the checked database behavior supports merging this change. Summary This PR keeps every tag available for permission checks when users filter or search secrets. - Secret listings check every tag before showing matching secrets. - Go integration tests can create roles and call the dashboard listing. Verified the real secret-list database code against disposable PostgreSQL using the same fixtures on the base and head revisions. On the base revision, filtering by allowed drops the restrict tag. On this branch, both normal and rank-paginated filtering preserve both tags, exclude unrelated secrets, and return no results fo
… the rest (313 characters)r a nonexistent tag. All four checks passed on this branch. No blocking issues were found. The change is safe to merge within the checked scope. HTTP permission enforcement and the full application test suite were not executed. Reviews (1) · Last reviewed commit: "regenerate api.gen" · Reviewed by the reviewer
#8670
2026-10-20 22:12
coderabbit
issue-comment
—summary
No actionable comments were generated in the recent review. 🎉 📝 Walkthrough The quick-add secret name input no longer uses explicit medium font styling or normal-weight placeholder styling. **Changes: Quick-add input styling** |Layer / File(s)|Summary| |:---|:---| |**Update secret name input classes** `frontend/src/pages/secret-manager/OverviewPage/components/QuickAddSecretRow/QuickAddSecretRow.tsx`|Removed the `font-medium` and `placeholder:font-normal` classes from the secret name input.| **Priority:** ⬇️ Low **Estimated code review effort:** 1 (Trivial) | ~2 minutes **Change:** Bug fix **Suggested reviewers:** `andrewhuhh` Merge Risk: ⚪ Minimal · up to efaf5 The quick-add inpu
… the rest (2312 characters)t now inherits the intended shared styling, with no identified merge-blocking risk. Pre-merge checks: 5 | Check name | Status | Explanation | | :------------------------ | :--------------------------------------------------------------------------------------------: | :-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | Title check | | The title clearly and concisely describes the UI fix and follows the conventional commit format. | | Description check | | The description explains the prior and current behavior, provides verification steps, identifies the change as a fix, and includes the relevant checklist items. The UI screenshot section and local-tes… | | Docstring Coverage | | No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check. Docstring coverage is scoped to functions touched by this diff. Analyzed 0 functions across 1… | | Linked Issues check | | Check skipped because no linked issues were found for this pull request. | | Out of Scope Changes check | | Check skipped because no linked issues were found for this pull request. | Full details: Description check **Explanation** The description explains the prior and current behavior, provides verification steps, identifies the change as a fix, and includes the relevant checklist items. The UI screenshot section and local-testing checkbox are not completed, but the description is otherwise mostly complete. Full details: Docstring Coverage **Explanation** No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check. Docstring coverage is scoped to functions touched by this diff. Analyzed 0 functions across 1 files. ✨ Finishing Touches 📝 Generate docstrings 🧪 Generate unit tests (beta) ### Usage-based review receipt - Mode: Continue automatically - Reviewed files: 1 - Waived: $0.25 (charged $0.00) - View usage details > [!NOTE] > This review exceeded your plan’s limits and used usage-based reviews—free during trial, billed after paid activation unless disabled. **Manage usage-based reviews**. Comment `@the reviewer help` to get the list of available commands.
#8670
2026-10-20 22:13
cubic
review-body
—summary
**No issues found** across 1 file View guided diff | Turn on auto-fix | Re-trigger the reviewer
#8670
2026-10-20 22:23
greptile
issue-comment
—summary
Confidence Score: 5/5 **[Low impact]** Safe to merge. Browser checks confirm the intended font-weight alignment without changing measured placeholder typography. Summary Removes `font-medium` and `placeholder:font-normal` from the quick-add secret name input so typed keys match existing secret keys. Browser checks using the actual input components, application CSS, and fonts confirmed that typed quick-add names change from font weight 500 to 400, matching existing-key presentations in both single- and multi-environment layouts. Placeholder weight remains 400, with the other measured placeholder styles unchanged. Verification used a focused rendering fixture with mocked project context,
… the rest (287 characters)not an authenticated full-app session. No actionable issues were found. This typography change is safe to merge. - Quick-add secret names now match the weight of existing rows. Reviews (1) · Last reviewed commit: "fix(ui): match quick-add secret key inpu..." · Reviewed by the reviewer
#8662
2026-10-20 22:26
coderabbit
issue-comment
—summary
📝 Walkthrough The secret input now tracks cursor and suggestion state while editing references. The reference wizard emits navigation strings and searches same-project secrets from the environment step, displaying results alongside filtered environments. **Changes: Secret reference editing and search** |Layer / File(s)|Summary| |:---|:---| |**Reference editing state** `frontend/src/components/v3/platform/SecretInput/InfisicalSecretInput.tsx`|The input tracks cursor position and suggestion activation, centralizes reference replacement, and controls popover visibility.| |**Wizard reference navigation** `frontend/src/components/v3/platform/SecretInput/SecretReferenceWizard.tsx`|The wizard a
… the rest (4593 characters)ccepts an `onNavigate` callback, synchronizes state with reference input, and emits reference strings for navigation.| |**Environment-step secret search** `frontend/src/components/v3/platform/SecretInput/SecretReferenceWizard.tsx`|The wizard searches same-project secrets, combines results with environment options, and supports selecting results and navigating breadcrumbs.| **Priority:** ➖ Normal **Estimated code review effort:** 3 (Moderate) | ~20 minutes **Change:** Bug fix ### Sequence Diagram(s) ```mermaid sequenceDiagram participant InfisicalSecretInput participant SecretReferenceWizard participant ProjectSecretQuickSearch InfisicalSecretInput->>SecretReferenceWizard: Pass reference input SecretReferenceWizard->>ProjectSecretQuickSearch: Search after debounce ProjectSecretQuickSearch-->>SecretReferenceWizard: Return matching secrets SecretReferenceWizard->>InfisicalSecretInput: Emit selected secret reference InfisicalSecretInput->>InfisicalSecretInput: Replace reference content ``` **Suggested reviewers:** `mathnogueira` Merge Risk: 🟡 Moderate · up to 731ac Adding a reference before another can silently remove part of a secret value. Fix the replacement boundary before merging. Security architecture risk: 🔵 Low · up to 731ac The new search remains restricted to the current project and uses existing server-side secret-description permissions. No introduced authorization bypass or secret-value disclosure was established. Some supporting response paths and downstream draft-saving behavior remain outside the verified scope. **Retained concerns** No architecture-level concerns identified. Security review details **Security Blast Radius** - _observed_ — The new wizard caller searches across environments and folders in the current project, not received cross-project grants. Its maximum discovery scope is therefore broader than the currently browsed folder, but ordinary-secret results remain subject to server-side permissions. **Trust Boundaries and Controls** - _observed_ — Client-supplied project and search parameters cross an authenticated server boundary. The server resolves permissions using the requested project together with actor identity, authentication method, and organization, then evaluates each ordinary secret's environment, path, key, and tags before returning it. **Hardening Proposals** - _proposed_ — Preserve cross-project identity directly from the selected grant, or defer final selection until source-project identity is available, rather than allowing an unqualified reference fallback. This addresses a dependency already present at the PR base, not an established regression from this change. Pre-merge checks: 5 | Check name | Status | Explanation | | :------------------------ | :--------------------------------------------------------------------------------------------: | :-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | Title check | | The title clearly and concisely describes the main UI fixes for selector scrolling, search, and dismissal. It follows the conventional commit format. | | Description check | | The description is complete. It explains the context, affected behavior, screenshots, verification steps, change type, checklist status, and frontend-only scope. | | Docstring Coverage | | No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check. Docstring coverage is scoped to functions touched by this diff. Analyzed 0 functions across 2… | | Linked Issues check | | Check skipped because no linked issues were found for this pull request. | | Out of Scope Changes check | | Check skipped because no linked issues were found for this pull request. | Full details: Docstring Coverage **Explanation** No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check. Docstring coverage is scoped to functions touched by this diff. Analyzed 0 functions across 2 files. ✨ Finishing Touches 📝 Generate docstrings 🧪 Generate unit tests (beta) ### Usage-based review receipt - Mode: Continue automatically - Reviewed files: 2 - Waived: $0.50 (charged $0.00) - View usage details > [!NOTE] > This review exceeded your plan’s limits and used usage-based reviews—free during trial, billed after paid activation unless disabled. **Manage usage-based reviews**. Comment `@the reviewer help` to get the list of available commands.
#8662
2026-10-20 22:30
greptile
issue-comment
—summary
Confidence Score: 4/5 **[Medium impact]** The changes affect the secret\-reference selector across value\-editing workflows\. The PR should not merge until the new heading satisfies the repository’s explicit UI styling requirement. Findings 1. **New heading breaks casing rules** Summary Updates the secret reference selector: - The reference selector follows the text and stays out of the way when it is not needed. - Typing a secret name finds matches across project environments and folders. One blocking repository requirement remains: remove the uppercase styling from the new “Secrets in this project” heading before merging. Reviews (1) · Last reviewed commit: "fix(ui): secret referen
… the rest (48 characters)ce selector scrol..." · Reviewed by the reviewer
#8662
2026-10-20 22:30
greptile
inline · P1
frontend/src/components/v3/platform/SecretInput/SecretReferenceWizard.tsx:584quality
**New heading breaks casing rules** The new “Secrets in this project” heading adds `uppercase`. The repository’s UI rule explicitly forbids casing overrides such as uppercase text in new UI, so this heading violates the required styling. Remove `uppercase` rather than copying the older headings’ styling. This requirement must be satisfied before merging. ```suggestion <div className="px-2 py-1.5 text-[10px] font-semibold tracking-wider text-muted"> Secrets in this project </div> ``` **Rule Used:** New UI must not override or deviate from the default styling of v3 design-system components (frontend/src/components/v3). Follow DESIGN.md. Flag: legacy v2
… the rest (235 characters)color tokens (e.g. mineshaft-*, bunker-*) instead of semantic tokens like bg-org or text-dange... (source) Note: If this suggestion doesn't match your team's coding style, reply to this and let me know. I'll remember it for next time!
#8662
2026-10-20 22:31
coderabbit
inline · P1
frontend/src/components/v3/platform/SecretInput/InfisicalSecretInput.tsx:92defect-data
**🗄️ Data Integrity & Integration** | **🟠 Major** | **⚡ Quick win** **Stop the right-boundary scan at the next reference, or navigation deletes user text.** `replaceReferenceContent` uses `getIndexOfUnclosedRefToTheRight` (Lines 24-29) to decide whether the current reference is already closed. The helper returns the first `}` anywhere to the right of the caret. It does not stop at whitespace or at a following `${`. Trigger: the value is `${dev. and ${prod.KEY}` and the caret is after `${dev.`. The helper finds the `}` of `${prod.KEY}`, so `isEnclosed` is true. `rhsValue` then starts at that `}`. Selecting an environment, folder, or secret replaces ` and ${prod.KEY` with the new content.
… the rest (883 characters) This change makes the problem more likely. `onNavigate` now calls `replaceReferenceContent` on every environment, folder, breadcrumb, and tab click, not only on the final selection. A common way to hit it is adding a new reference in front of an existing one, for example in `postgres://${user}:${pass}@...`. The text is removed without any warning and can be saved with the secret. Fix the boundary in the helper so that a reference ends only at its own `}`. 🐛 Proposed fix for the boundary helper ```diff const getIndexOfUnclosedRefToTheRight = (value: string, pos: number) => { for (let i = pos; i < value.length; i += 1) { if (value[i] === "}") return i - 1; + // A new reference or whitespace ends the current (unclosed) reference + if (value[i] === "$" && value[i + 1] === "{") return -1; + if (/\s/.test(value[i])) return -1; } return -1; }; ```
#8662
2026-10-20 22:31
coderabbit
review-body
—summary
**Actionable comments posted: 1**
#8662
2026-10-20 22:34
cubic
inline · P1
frontend/src/components/v3/platform/SecretInput/InfisicalSecretInput.tsx:282defect-data
Stop the right-side scan at the next `${` or whitespace before applying navigation updates; otherwise a later `}` can make this call delete intervening user text.
#8662
2026-10-20 22:34
cubic
inline · P2
frontend/src/components/v3/platform/SecretInput/SecretReferenceWizard.tsx:621defect-functional
A failed deep-search request is shown as “No matching environments or secrets,” falsely implying the search succeeded. Read the query error state and show an error message instead.
#8662
2026-10-20 22:34
cubic
review-body
—summary
**2 issues found** across 2 files Reply with feedback, questions, or to request a fix.View guided diff | Turn on auto-fix | Re-trigger the reviewer
#7670
2026-10-20 22:36
coderabbit
issue-comment
—summary
📝 Walkthrough This change adds Portainer as an app connection and secret-sync destination. It adds backend connection and synchronization operations, frontend connection and destination forms, and Portainer integration and API documentation. **Changes: Portainer app connections** |Layer / File(s)|Summary| |:---|:---| |**Connection contracts and API** `backend/src/services/app-connection/portainer/*`, `backend/src/services/app-connection/app-connection-*`, `backend/src/server/routes/v1/app-connection-routers/*`|Adds API-token connection schemas and validation, Portainer request handling through direct or gateway routes, environment and stack listing, and app-connection routes and service
… the rest (9664 characters)registration.| |**Connection setup and discovery** `frontend/src/pages/organization/AppConnections/...`, `frontend/src/hooks/api/appConnections/portainer/*`, `frontend/src/helpers/appConnections.ts`|Adds the Portainer connection form and query hooks for environments and stacks. The form supports API-token credentials and optional gateway selection.| **Portainer secret sync** |Layer / File(s)|Summary| |:---|:---| |**Synchronization behavior and API** `backend/src/services/secret-sync/portainer/*`, `backend/src/services/secret-sync/secret-sync-*`, `backend/src/server/routes/v1/secret-sync-routers/*`|Adds Portainer destination schemas and routes, duplicate detection, and secret retrieval, synchronization, import, and removal. Compose values are encoded and decoded; Kubernetes stacks are excluded. Tests cover supported stack types and synchronization cases.| |**Destination setup and display** `frontend/src/components/secret-syncs/forms/*`, `frontend/src/pages/secret-manager/...`, `frontend/src/hooks/api/secretSyncs/*`, `frontend/src/helpers/secretSyncs.ts`|Adds Portainer destination validation, environment and stack selection, review and detail displays, and integration-table destination values.| |**Portainer documentation and API reference** `backend/src/lib/api-docs/constants.ts`, `docs/api-reference/endpoints/{app-connections,secret-syncs}/portainer/*`, `docs/integrations/*/portainer.mdx`, `docs/snippets/*`, `docs/docs.json`|Adds Portainer connection and secret-sync guides, endpoint metadata, navigation and catalog entries, form and token-generation snippets, and API documentation fields.| **Priority:** ➖ Normal **Estimated code review effort:** 4 (Complex) | ~60 minutes **Change:** Feature ### Sequence Diagram(s) ```mermaid sequenceDiagram participant User participant PortainerSyncFields participant AppConnectionRoute participant PortainerConnectionService participant PortainerAPI User->>PortainerSyncFields: Select connection PortainerSyncFields->>AppConnectionRoute: Request environments and stacks AppConnectionRoute->>PortainerConnectionService: List environments and stacks PortainerConnectionService->>PortainerAPI: Request environments and stacks PortainerAPI-->>PortainerConnectionService: Return environment and stack data PortainerConnectionService-->>PortainerSyncFields: Return mapped results PortainerSyncFields-->>User: Show environment and stack options ``` ```mermaid sequenceDiagram participant SecretSyncRoute participant SecretSyncFns participant PortainerSyncFns participant PortainerAPI SecretSyncRoute->>SecretSyncFns: Dispatch Portainer sync SecretSyncFns->>PortainerSyncFns: Sync secrets for selected stack PortainerSyncFns->>PortainerAPI: Read stack and update environment PortainerAPI-->>PortainerSyncFns: Return stack and update response PortainerSyncFns-->>SecretSyncFns: Return sync result SecretSyncFns-->>SecretSyncRoute: Return operation result ``` Merge Risk: 🔵 Low · up to b66d0 This change adds Portainer connections and secret syncing to stack environment variables. Two minor issues remain. If a connection uses a plain http:// Portainer URL without a Gateway, the API token is sent unencrypted. An empty environment or stack selection also shows a generic validation message. Neither issue blocks normal use, so the change is mergeable once the owner accepts or addresses them. Security architecture risk: 🟡 Moderate · up to b66d0 HTTP configurations can expose Portainer tokens and synchronized secrets. Concurrent writers to the same stack can also undo secret removal or rotation. Authentication and destination checks limit exposure, but recovery after an interrupted deployment remains uncertain. **Retained concerns** - **Low · security · observed:** The new Portainer connection accepts HTTP URLs and forwards its API token and synchronization payloads without requiring TLS. A network observer or active intermediary on a plaintext segment can obtain credentials and secrets. HTTPS configurations verify certificates, and gateway tunneling protects its tunnel, but neither requires encryption on an HTTP destination hop. - **Medium · security · inferred:** Portainer synchronization and removal replace the entire environment-variable list without destination-scoped serialization or a conditional-write token. When separate syncs share a stack, one can retain another's old values and write them back after successful removal or rotation. Per-sync locks do not prevent this interleaving. Duplicate-target detection is useful counterevidence, but enforcement is optional. Successful cleanup can delete its sync record, leaving no subsequent reconciliation from that owner. Security review details **Security Blast Radius** - _inferred_ — Transport exposure includes secrets carried by the configured sync and the connection's API token. A stolen token may authorize resources beyond the selected stack; its actual Portainer privileges are not established. Concurrent-write effects are bounded to writers sharing the same external stack. No cross-tenant escalation was established. **Security Findings and Attack Paths** - _observed_ — The retained transport finding is reachable when an authorized configurator selects an HTTP instance URL. Subsequent requests send X-API-Key and applicable secret-bearing bodies over that protocol. Exploitation requires access to the plaintext network segment; anonymous callers cannot simply bypass the connection authorization boundary. - _inferred_ — A writer can read an old list, retain another owner's secret, and submit that stale value after the other owner removes or rotates it. Both operations can complete without an overlapping deployment conflict. Per-sync locking and fresh reads on retries do not reconcile a successful stale replacement; optional duplicate-target blocking reduces, but does not universally prevent, this exposure. **Trust Boundaries and Controls** - _observed_ — Discovery routes require JWT or OAuth authentication and obtain the connection through the shared Connect authorization path. That path rejects another organization's connection and checks connection-specific project or organization permission. Sync usage additionally rejects a project-scoped connection belonging to a different project. - _observed_ — Direct requests inherit validated-address pinning, private-address restrictions subject to existing configuration, and redirect blocking. Gateway attachment is organization-scoped during connection creation, and gateway routing uses target host/port credentials. These controls constrain destination reachability but do not enforce HTTPS for Portainer. **Resilience and Maintainability Implications** - _inferred_ — If Portainer persists Env before a deployment fails or is interrupted, a retry's equality shortcut cannot distinguish stored configuration from successful workload rollout. The adapter awaits updates and propagates errors, but has no deployment-status reconciliation or rollback. Whether supported Portainer versions exhibit this persistence ordering remains an external-contract gap, not a verified failure. **Hardening Proposals** - _proposed_ — Require HTTPS for credential-bearing Portainer destinations, including the gateway-to-Portainer hop. - _proposed_ — Enforce exclusive ownership or serialize the complete read-modify-write operation by canonical Portainer destination identity. Where external writers remain possible, use a supported conditional-update or conflict-reconciliation contract. Establish deployment recovery semantics before treating stored-variable equality as proof of completed rollout. Pre-merge checks: 5 | Check name | Status | Explanation | | :------------------------ | :--------------------------------------------------------------------------------------------: | :-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | Title check | | The title clearly identifies the feature and follows the conventional commit format. It accurately summarizes the Portainer app-connection and secret-sync support. | | Description check | | The description is complete and relevant. It includes context, a screenshot, verification steps, feature classification, checklist status, implementation details, and testing information. | | Docstring Coverage | | No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check. Docstring coverage is scoped to functions touched by this diff. Analyzed 0 functions across 5… | | Linked Issues check | | Check skipped because no linked issues were found for this pull request. | | Out of Scope Changes check | | Check skipped because no linked issues were found for this pull request. | Full details: Docstring Coverage **Explanation** No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check. Docstring coverage is scoped to functions touched by this diff. Analyzed 0 functions across 50 files. (29 skipped: 22 unsupported, 7 over the file limit.) ✨ Finishing Touches 📝 Generate docstrings 🧪 Generate unit tests (beta) ### Usage-based review receipt - Mode: Continue automatically - Reviewed files: 79 - Waived: $19.75 (charged $0.00) - View usage details > [!NOTE] > This review exceeded your plan’s limits and used usage-based reviews—free during trial, billed after paid activation unless disabled. **Manage usage-based reviews**. Comment `@the reviewer help` to get the list of available commands.
#7670
2026-10-20 22:44
coderabbit
inline · P2
backend/src/services/app-connection/portainer/portainer-connection-schemas.ts:21defect-security
**🔒 Security & Privacy** | **🛡️ Detected with Advanced Tier** | **🟡 Minor** | **⚡ Quick win** **Sensitive Data Exposure** **Reachability:** External **Exploitability:** Difficult **CWE:** CWE-319 — Cleartext Transmission of Sensitive Information **Require HTTPS for direct Portainer requests.** The schema accepts `http://` URLs, and `safeRequest` sends the `X-API-Key` header through an `http.Agent` on the direct path. Reject non-HTTPS URLs when no Gateway is configured. Enforce HTTPS for direct requests ```diff + if (!gatewayId && url.protocol !== "https:") { + throw new BadRequestError({ + message: "Portainer instance URL must use HTTPS unless the connection is routed throug
… the rest (75 characters)h a Gateway." + }); + } + if (!gatewayId || !gatewayV2Service) { ```
#7670
2026-10-20 22:44
coderabbit
inline · P2
frontend/src/components/secret-syncs/forms/schemas/portainer-sync-destination-schema.ts:17defect-functional
**🎯 Functional Correctness** | **🟡 Minor** | **⚡ Quick win** **The "Environment is required" and "Stack is required" messages never show for an empty selection.** `PortainerSyncFields` sets both fields to `undefined` when the connection or environment changes. In Zod 3, an `undefined` input uses `required_error`, not `invalid_type_error`. If `required_error` is not set, Zod uses the default message "Required". So an empty Environment or Stack field shows "Required" and not the intended text. Set `required_error` on both fields. 🐛 Proposed fix ```diff --- "a/frontend/src/components/secret-syncs/forms/schemas/portainer-sync-destination-schema.ts" +++ "b/frontend/src/components/secret-sy
… the rest (1484 characters)ncs/forms/schemas/portainer-sync-destination-schema.ts" @@ -7,14 +7,20 @@ z.object({ destination: z.literal(SecretSync.Portainer), destinationConfig: z.object({ environmentId: z - .number({ invalid_type_error: "Environment is required" }) + .number({ + required_error: "Environment is required", + invalid_type_error: "Environment is required" + }) .int() .positive("Environment is required"), stackId: z - .number({ invalid_type_error: "Stack is required" }) + .number({ + required_error: "Stack is required", + invalid_type_error: "Stack is required" + }) .int() .positive("Stack is required") }) }) ); ``` 📝 Committable suggestion > ‼️ **IMPORTANT** > Carefully review the code before committing. Ensure that it accurately replaces the highlighted code, contains no missing lines, and has no issues with indentation. Thoroughly test & benchmark the code to ensure it meets the requirements. ```suggestion environmentId: z .number({ required_error: "Environment is required", invalid_type_error: "Environment is required" }) .int() .positive("Environment is required"), stackId: z .number({ required_error: "Stack is required", invalid_type_error: "Stack is required" }) .int() .positive("Stack is required") ```
#7670
2026-10-20 22:44
coderabbit
review-body
—summary
**Actionable comments posted: 2**
#7670
2026-10-20 22:46
greptile
issue-comment
—summary
Confidence Score: 3/5 **[High impact]** The feature sends selected Infisical secrets into Portainer stacks and can trigger stack redeploys\. Not safe to merge until both Gateway reliability issues are fixed, because valid Portainer secret syncs can fail during deployment or stop retrying temporary conflicts. Findings 1. **Long deployments lose their connection** 2. **Temporary conflicts stop retries** Summary Adds Portainer app connections and secret syncs for Compose and Swarm stacks, including API tokens, Gateways, Gateway pools, environment and stack pickers, import and removal support, tests, and setup guides. Two Gateway reliability issues need fixing before merging: - Portainer
… the rest (563 characters)connections can reach instances and list supported stacks. - Portainer syncs write and import stack environment variables. - The sync wizard lets teams choose Portainer environments and stacks. Intentional behavior acknowledged by PierreOucif: destinations are distinguished by Gateway route to allow overlapping instance-local addresses. PierreOucif's guide also states that the sync does not check deployment results after the update request returns. Reviews (1) · Last reviewed commit: "docs(secret-sync): drop the duplicated p..." · Reviewed by the reviewer
#7670
2026-10-20 22:46
greptile
inline · P1
backend/src/services/app-connection/portainer/portainer-connection-fns.ts:175defect-functional
**Long deployments lose their connection** If Portainer takes more than two minutes to redeploy a file-based stack before replying, this Gateway request loses its connection even though the update allows four minutes. The call leaves the Gateway's two-minute idle timeout enabled. The connection closes while deployment is still running, and the sync can ultimately fail despite Portainer completing the deployment. Request retries can delay the final failure, but do not prevent the premature disconnect. Enable `longLived` for these longer Gateway requests so the bounded request timeout controls the wait. This needs fixing before merging. Artifacts **Executed Portainer Gateway timeout test**
… the rest (820 characters) - The authored script executes the real sync and Gateway helpers against a local TLS fixture, explicitly showing the stubs, time scaling, and in-memory fix. **Fixture startup log before ALPN correction** - The initial command failed TLS protocol negotiation before reaching an endpoint, documenting a fixture setup failure rather than candidate evidence. **Current code disconnects during a quiet update** - The before command captures real requests, Gateway timeout teardown, deployment-time disconnect, retry, and sync failure, confirming the premature connection loss. **In-memory fix receives the completed update** - The after command repeats the same fixture with longLived enabled and receives HTTP 200 OK, confirming the idle-timeout bypass resolves this reproduction. Ran code and verified through TREX
#7670
2026-10-20 22:46
greptile
inline · P1
backend/src/services/secret-sync/portainer/portainer-sync-fns.ts:117defect-functional
**Temporary conflicts stop retries** When Portainer returns `409` because a deployment is already running, a Gateway-backed sync stops retrying instead of allowing the temporary conflict to clear. `withGatewayV2Proxy` wraps the upstream HTTP error in `BadRequestError`, losing its HTTP status. It then misses the retryable branch in `toStackUpdateError`, and this blanket classification sets `shouldRetry: false`. The queue ends the attempt sequence, leaving the secrets unsynced until another sync runs. The identical response through a direct connection remains retryable. Preserve the upstream error status or distinguish temporary upstream failures from configuration errors so Gateway syncs re
… the rest (89 characters)tain their retries. This needs fixing before merging. Ran code and verified through TREX
#7670
2026-10-20 22:47
cubic
inline · P1
frontend/src/components/secret-syncs/forms/SecretSyncDestinationFields/PortainerSyncFields.tsx:55defect-functional
Creating a Portainer connection saves the form before this callback clears the destination IDs, so OAuth restoration pairs the new connection with the old IDs. Those positive IDs pass validation and can target the wrong stack; clear them before storing the form snapshot.
#7670
2026-10-20 22:47
cubic
inline · P1
backend/src/services/secret-sync/portainer/portainer-sync-fns.ts:174defect-functional
Portainer 2.45+ Git stacks may expose `GitSourceId` without `GitConfig`, so this guard fails the sync before the existing `sourceID` update path can run. Allow source-backed stacks to build their payload without legacy `GitConfig`.
#7670
2026-10-20 22:47
cubic
inline · P1
backend/src/services/secret-sync/secret-sync-fns.ts:373defect-functional
Git-backed Portainer syncs cannot apply changed secrets because `writeGitStackEnv` calls Portainer's `/stacks/{id}/git` update endpoint with `POST`; use `PUT` for that endpoint.
#7670
2026-10-20 22:47
cubic
inline · P1
backend/src/services/secret-sync/portainer/portainer-sync-fns.ts:98defect-functional
Legacy Portainer Git updates omit `repositoryURL`, so the POST replaces the stack’s repository URL with an empty value. Preserve `gitConfig.URL` in the non-`sourceID` payload; otherwise later deployments cannot clone the stack.
#7670
2026-10-20 22:47
cubic
inline · P1
backend/src/services/app-connection/portainer/portainer-connection-fns.ts:130defect-security
Reject non-HTTPS URLs before making this direct request; it sends the `X-API-Key` header over cleartext HTTP.
#7670
2026-10-20 22:47
cubic
inline · P2
backend/src/services/secret-sync/portainer/portainer-sync-fns.ts:99defect-functional
For legacy Git stacks using inline credentials, this payload never sends `auth.Password`, so the update can discard credentials and break the next deployment. Include `repositoryPassword` when the returned password is present. ```suggestion ...(auth?.Password !== undefined ? { repositoryPassword: auth.Password } : {}), repositoryUsername: auth?.Username ?? "", ```
#7670
2026-10-20 22:47
cubic
inline · P2
backend/src/services/secret-sync/portainer/portainer-sync-fns.ts:92defect-functional
Git updates omit `configFilePath`, so stacks using a non-default compose file lose their compose path when this replacement payload is saved. Include `gitConfig.ConfigFilePath` in the payload. ```suggestion repositoryReferenceName: gitConfig.ReferenceName ?? "", configFilePath: gitConfig.ConfigFilePath ?? "", ```
#7670
2026-10-20 22:47
cubic
inline · P3
docs/integrations/secret-syncs/portainer.mdx:80docs
The initial-sync section omits that enabling **Include secrets from all subfolders** makes both import options invalid. Add that users must choose **Overwrite Portainer** in this case; otherwise API users can follow the documented combination and receive validation errors. ```suggestion Under **Initial Sync Behavior**, choose how Infisical handles variables that already exist on the stack on the first sync. If **Include secrets from all subfolders** is enabled, choose **Overwrite Portainer** because the import options aren't available: ```
#7670
2026-10-20 22:47
cubic
review-body
—summary
**9 issues found** across 91 files Reply with feedback, questions, or to request a fix.View guided diff | Turn on auto-fix | Re-trigger the reviewer
#7670
2026-10-20 22:47
cubic
inline · P3
frontend/src/components/secret-syncs/forms/schemas/portainer-sync-destination-schema.ts:11defect-functional
Set `required_error` on both number schemas; clearing either selection currently shows Zod’s generic “Required” instead of the field-specific message.
#8669
2026-10-20 22:51
coderabbit
issue-comment
—summary
📝 Walkthrough The changes add private-key guidance for Palo Alto Networks certificate syncs. The create and edit forms display the guidance for Palo Alto Networks destinations. The documentation describes private-key prerequisites and certificate handling. **Changes: Palo Alto Networks guidance** |Layer / File(s)|Summary| |:---|:---| |**Define and surface private-key guidance** `frontend/src/helpers/pkiSyncs.ts`, `frontend/src/components/pki-syncs/forms/CreatePkiSyncForm.tsx`, `frontend/src/components/pki-syncs/forms/EditPkiSyncForm.tsx`, `docs/documentation/platform/pki/applications/certificate-syncs/palo-alto-networks.mdx`, `docs/documentation/platform/pki/applications/certificate-sync
… the rest (2741 characters)s/palo-alto-networks-ssl-tls-profile.mdx`|Adds shared guidance that certificates without private keys are skipped. The create and edit forms append the guidance for Palo Alto Networks destinations. The documentation describes private-key prerequisites, including the exception for imported issuing CA certificates.| **Priority:** ⬇️ Low **Estimated code review effort:** 2 (Simple) | ~8 minutes **Change:** Other **Suggested reviewers:** `carlosmonastyrski` Merge Risk: 🔵 Low · up to ad3ac General Palo Alto Networks users may incorrectly treat valid issuing CA certificates as unusable because the form gives them an overly broad private-key requirement. Pre-merge checks: 5 | Check name | Status | Explanation | | :------------------------ | :--------------------------------------------------------------------------------------------: | :-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | Description check | | The description clearly explains the private-key requirement, the skip behavior, and the Palo Alto Networks context. It includes the required Context and Type sections, and the checklist is mostly com… | | Title check | | The title clearly and concisely describes the main change: documenting that Palo Alto Networks syncs require certificate private keys. It uses a valid conventional commit format. | | Docstring Coverage | | No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check. Docstring coverage is scoped to functions touched by this diff. Analyzed 0 functions across 3… | | Linked Issues check | | Check skipped because no linked issues were found for this pull request. | | Out of Scope Changes check | | Check skipped because no linked issues were found for this pull request. | Full details: Description check **Explanation** The description clearly explains the private-key requirement, the skip behavior, and the Palo Alto Networks context. It includes the required Context and Type sections, and the checklist is mostly complete. The Screenshots and Steps to verify sections are empty, but these omissions are non-critical for this documentation-focused change. Full details: Docstring Coverage **Explanation** No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check. Docstring coverage is scoped to functions touched by this diff. Analyzed 0 functions across 3 files. (2 skipped: 2 unsupported.) ✨ Finishing Touches 📝 Generate docstrings 🧪 Generate unit tests (beta) Comment `@the reviewer help` to get the list of available commands.
#8669
2026-10-20 22:52
cubic
inline · P2
docs/documentation/platform/pki/applications/certificate-syncs/palo-alto-networks-ssl-tls-profile.mdx:21docs
This makes the CSR's origin sound like the blocker, so users who import the matching private key are told their certificate will be skipped. State that the sync skips certificates only when their private key is unavailable in Infisical. ```suggestion An SSL/TLS service profile can only use a certificate that was imported with its private key. If a certificate was issued from a CSR generated elsewhere and its private key isn't available in Infisical, the sync can't deploy it and skips it. See [Imported Certificate Not Showing Under SSL/TLS Service Profile](https://knowledgebase.paloaltonetworks.com/KCSArticleDetail?id=kA10g000000PLmJCAW). ```
#8669
2026-10-20 22:52
cubic
inline · P2
frontend/src/components/pki-syncs/forms/CreatePkiSyncForm.tsx:384defect-functional
This adds the warning only to the `lg:flex` sidebar, which is hidden on smaller viewports, so mobile users never see it. Render the note in the certificate step's main content or otherwise make it visible responsively.
#8669
2026-10-20 22:52
cubic
review-body
—summary
**2 issues found** across 5 files Reply with feedback, questions, or to request a fix.View guided diff | Turn on auto-fix | Re-trigger the reviewer
#8669
2026-10-20 22:55
coderabbit
inline · P2
frontend/src/helpers/pkiSyncs.ts:205defect-functional
**🎯 Functional Correctness** | **🟡 Minor** | **⚡ Quick win** **Scope the private-key requirement to deployed certificates.** The general Palo Alto Networks sync can import issuing CA certificates without private keys. The shared note currently says every certificate needs a private key, so both create and edit forms show incorrect guidance. Update the shared constant; both forms consume it. Suggested fix ```diff --- "a/frontend/src/helpers/pkiSyncs.ts" +++ "b/frontend/src/helpers/pkiSyncs.ts" @@ -201,8 +201,8 @@ PkiSync.PaloAltoNetworksSslTlsProfile ]; export const PALO_ALTO_NETWORKS_PRIVATE_KEY_NOTE = - "Palo Alto Networks needs each certificate's private key. Certificates wit
… the rest (1095 characters)hout one, such as those issued from a CSR generated outside Infisical, are skipped."; + "Palo Alto Networks needs each deployed certificate's private key. Issuing CA certificates imported alongside do not need one. Other certificates without one, such as those issued from a CSR generated outside Infisical, are skipped."; export const isPaloAltoNetworksPkiSync = (destination?: PkiSync) => destination === PkiSync.PaloAltoNetworks || destination === PkiSync.PaloAltoNetworksSslTlsProfile; ``` 📝 Committable suggestion > ‼️ **IMPORTANT** > Carefully review the code before committing. Ensure that it accurately replaces the highlighted code, contains no missing lines, and has no issues with indentation. Thoroughly test & benchmark the code to ensure it meets the requirements. ```suggestion export const PALO_ALTO_NETWORKS_PRIVATE_KEY_NOTE = "Palo Alto Networks needs each deployed certificate's private key. Issuing CA certificates imported alongside do not need one. Other certificates without one, such as those issued from a CSR generated outside Infisical, are skipped."; ```
#8669
2026-10-20 22:55
coderabbit
review-body
—summary
**Actionable comments posted: 1**
#8669
2026-10-20 23:02
greptile
issue-comment
—summary
Confidence Score: 5/5 **[Low impact]** Safe to merge. The changes add documentation and destination-specific guidance without changing certificate syncing behavior. Summary Documents that Palo Alto Networks certificate syncs need the certificate's private key. - Palo Alto sync forms explain when certificates are skipped. No actionable issues were found. The change is safe to merge. Reviews (1) · Last reviewed commit: "chore: Add PAN private key warning on UI..." · Reviewed by the reviewer
#8676
2026-10-20 23:03
coderabbit
issue-comment
—summary
No actionable comments were generated in the recent review. 🎉 📝 Walkthrough The organization/project switcher now uses a responsive maximum width in project scope. It hides organization badges in project scope and uses shorter badge labels elsewhere. **Changes: Switcher presentation** |Layer / File(s)|Summary| |:---|:---| |**Scope-based switcher presentation** `frontend/src/layouts/OrganizationLayout/components/NavBar/Navbar.tsx`|In project scope, the switcher uses a maximum width of `max-w-52` and omits organization badges. In other scopes, badges remain visible with the labels “Org” and “Sub.”| **Priority:** ⬇️ Low **Estimated code review effort:** 1 (Trivial) | ~5 minutes **Chang
… the rest (1735 characters)e:** Bug fix **Suggested reviewers:** `scott-ray-wilson` Merge Risk: ⚪ Minimal · up to 49443 The scoped UI presentation changes match the stated objectives and are ready to merge. Pre-merge checks: 5 | Check name | Status | Explanation | | :------------------------ | :--------------------------------------------------------------------------------------------: | :-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | Title check | | The title clearly describes the main UI change and follows the conventional commit format. | | Description check | | The description is complete and matches the repository template. It explains the prior and new behavior, includes screenshots, verification steps, change type, and checklist status. | | Docstring Coverage | | No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check. Docstring coverage is scoped to functions touched by this diff. Analyzed 0 functions across 1… | | Linked Issues check | | Check skipped because no linked issues were found for this pull request. | | Out of Scope Changes check | | Check skipped because no linked issues were found for this pull request. | Full details: Docstring Coverage **Explanation** No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check. Docstring coverage is scoped to functions touched by this diff. Analyzed 0 functions across 1 files. ✨ Finishing Touches 📝 Generate docstrings 🧪 Generate unit tests (beta) Comment `@the reviewer help` to get the list of available commands.
#8676
2026-10-20 23:04
cubic
review-body
—summary
**No issues found** across 1 file View guided diff | Turn on auto-fix | Re-trigger the reviewer
#8676
2026-10-20 23:14
greptile
issue-comment
—summary
Confidence Score: 5/5 **[Low impact]** Safe to merge. No blocking issues or new layout regressions were established. Summary Keeps the org or sub-org name visible in project and product scope by replacing the fixed 72px org cell with `max-w-52`. Badges are hidden in those scopes, and organization-scope labels are shortened to “Org” and “Sub”. Browser checks covered 98 before-and-after cases across organization, product, and project scope, including root organizations, sub-organizations, long names, and three viewport widths. Badge assertions passed, and no new layout regression was established. These checks used the actual Navbar and frontend styles with mocked service dependencies, rath
… the rest (253 characters)er than an authenticated deployment. No actionable issues found. Safe to merge. - The header shows the org name in project and product scope. Reviews (1) · Last reviewed commit: "fix(ui): keep the org name visible in pr..." · Reviewed by the reviewer
#8675
2026-10-20 23:16
coderabbit
issue-comment
—summary
No actionable comments were generated in the recent review. 🎉 📝 Walkthrough Sheet and Dialog footers were updated across the frontend to use right alignment and consistent action ordering. The changes also add footer guidance, update Sheet stories, and adjust organization identity creation button variants. **Changes: Footer alignment and action ordering** |Layer / File(s)|Summary| |:---|:---| |**Apply footer layout conventions** `DESIGN.md`, `frontend/src/components/v3/generic/Sheet/Sheet.stories.tsx`, `frontend/src/pages/cert-manager/.../CertificateTemplateEnrollmentModal.tsx`, `frontend/src/pages/project/RoleDetailsBySlugPage/components/*`, `frontend/src/pages/project/components/Fold
… the rest (3756 characters)erAccessSection/EditFolderAccessSheet.tsx`, `frontend/src/pages/secret-manager/OverviewPage/components/SecretDropzone/*`, `frontend/src/pages/secret-manager/SecretDashboardPage/.../VaultSecretImportModal.tsx`|Added footer layout guidance and applied right alignment in Sheet stories and selected certificate, role, folder, and secret-management views. The CSV mapping dialog keeps Back on the left.| |**Update identity and role form actions** `frontend/src/pages/organization/AccessManagementPage/.../IdentitySection/*`, `frontend/src/pages/project/AccessControlPage/.../IdentityTab/...`, `frontend/src/pages/project/IdentityDetailsByIDPage/.../IdentityRoleModify.tsx`, `frontend/src/pages/project/MemberDetailsByIDPage/.../MemberMultiRoleModify.tsx`|Right-aligned form footers and moved Cancel before the primary action. Organization identity creation selects its button variant using `isSubOrganization`.| |**Update approval and settings actions** `frontend/src/pages/organization/ProductSettingsPage/.../HoneyTokenModal.tsx`, `frontend/src/pages/secret-manager/SecretApprovalsPage/components/*`, `frontend/src/pages/secret-manager/SettingsPage/.../SecretValidationRulesSection.tsx`, `frontend/src/views/Alerts/AlertForm.tsx`|Right-aligned approval, policy, validation, alert, and honey-token actions. Several forms now place Cancel or Reject before the primary action.| **Priority:** ⬇️ Low **Estimated code review effort:** 2 (Simple) | ~12 minutes **Change:** Bug fix **Suggested reviewers:** `andrewhuhh`, `scott-ray-wilson` Merge Risk: ⚪ Minimal · up to 19de7 The footer changes have no identified material regression and are mergeable after normal checks. Pre-merge checks: 5 | Check name | Status | Explanation | | :------------------------ | :--------------------------------------------------------------------------------------------: | :-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | Title check | | The title clearly summarizes the primary UI change and follows the conventional commit format. | | Description check | | The description includes the required context, screenshots, verification steps, change type, and checklist. It clearly documents the scope and behavior of the footer standardization. | | Docstring Coverage | | No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check. Docstring coverage is scoped to functions touched by this diff. Analyzed 0 functions across 2… | | Linked Issues check | | Check skipped because no linked issues were found for this pull request. | | Out of Scope Changes check | | Check skipped because no linked issues were found for this pull request. | Full details: Docstring Coverage **Explanation** No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check. Docstring coverage is scoped to functions touched by this diff. Analyzed 0 functions across 23 files. (1 skipped: 1 unsupported.) ✨ Finishing Touches 📝 Generate docstrings 🧪 Generate unit tests (beta) ### Usage-based review receipt - Mode: Continue automatically - Reviewed files: 24 - Waived: $6.00 (charged $0.00) - View usage details > [!NOTE] > This review exceeded your plan’s limits and used usage-based reviews—free during trial, billed after paid activation unless disabled. **Manage usage-based reviews**. > [!WARNING] > ⚠️ This pull request shows signs of AI-generated slop (description_diff_mismatch). It has been flagged by the reviewer slop detection and should be reviewed carefully. Comment `@the reviewer help` to get the list of available commands.
#8675
2026-10-20 23:23
cubic
review-body
—summary
**No issues found** across 24 files View guided diff | Turn on auto-fix | Re-trigger the reviewer
#8675
2026-10-20 23:24
greptile
issue-comment
—summary
Confidence Score: 5/5 **[Medium impact]** Safe to merge. The changes preserve button handlers, form associations, and submit types, while the tested Sheet story retains accessible actions and working dismissal at desktop and mobile widths. Summary Standardizes sheet footers without changing their submit or cancel behavior. - Sheet footers put their actions on the right, with Cancel before the main action. - Organization identity creation matches its button style to the organization scope. - Access review choices sit at the right, with rejection before approval. Browser checks of the updated Sheet story confirmed right-aligned actions without overlap or overflow at 1440px, 375px, and 320p
… the rest (271 characters)x viewport widths. Close continued to dismiss the sheet, and the footer remained anchored while its content scrolled. No actionable issues found. Safe to merge. Reviews (1) · Last reviewed commit: "fix(ui): right-align sheet footers with ..." · Reviewed by the reviewer
#8660
2026-10-20 23:26
coderabbit
issue-comment
—summary
No actionable comments were generated in the recent review. 🎉 📝 Walkthrough The agent reference adds the `file-output.combine-certificate-chain` option and documents how combined certificate-and-chain output interacts with chain paths, root omission, reloads, and restarts. **Changes: Full-chain output documentation** |Layer / File(s)|Summary| |:---|:---| |**Document combined certificate-and-chain output** `docs/documentation/platform/pki/reference/infisical-agent.mdx`|The reference describes the new option, which writes the certificate and chain to `certificate.path`. It documents that `chain.path` is ignored when the option is enabled, while `chain.omit-root` remains effective. It als
… the rest (2878 characters)o covers reload and restart behavior.| **Priority:** ⬇️ Low **Estimated code review effort:** 1 (Trivial) | ~4 minutes **Change:** Other **Suggested reviewers:** `adkah` Merge Risk: ⚪ Minimal · up to bd6a1 The reference explains how to reload NGINX after certificate renewal. No actionable issue remains before merge. Architecture risk: 🔵 Low · up to bd6a1 The change affects 1 system. **Changed systems:** `docs` **Architecture concerns** No architecture-level concerns identified. Review details **Systems and components** - _observed_ — docs (service) was modified; 1 changed file maps to changed impact. **Before / after behavior** - _observed_ — Modified behavior in docs/documentation/platform/pki/reference/infisical-agent.mdx: Added the `file-output.combine-certificate-chain` configuration option, which writes the certificate followed by its chain to `certificate.path`; when enabled, `chain.path` is ignored while `chain.omit-root` remains effective. - _observed_ — Modified behavior in docs/documentation/platform/pki/reference/infisical-agent.mdx: Added a full-chain certificate file section documenting combined certificate-and-chain output, including configuration, server integration, post-event reloads, root CA inclusion control, and restart-time rewrite and issuance-hook behavior. Pre-merge checks: 5 | Check name | Status | Explanation | | :------------------------ | :--------------------------------------------------------------------------------------------: | :-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | Title check | | The title clearly describes the documentation change and follows the conventional commit format. | | Description check | | The description includes context, verification steps, change type, and checklist information. It accurately identifies the documentation-only scope and dependency on the CLI release. | | Docstring Coverage | | No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check. Docstring coverage is scoped to functions touched by this diff. Analyzed 0 functions across 0… | | Linked Issues check | | Check skipped because no linked issues were found for this pull request. | | Out of Scope Changes check | | Check skipped because no linked issues were found for this pull request. | Full details: Docstring Coverage **Explanation** No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check. Docstring coverage is scoped to functions touched by this diff. Analyzed 0 functions across 0 files. (1 skipped: 1 unsupported.) ✨ Finishing Touches 🧪 Generate unit tests (beta) Comment `@the reviewer help` to get the list of available commands.
#8660
2026-10-20 23:27
cubic
review-body
—summary
**No issues found** across 1 file View guided diff | Turn on auto-fix | Re-trigger the reviewer
#8660
2026-10-20 23:31
greptile
issue-comment
—summary
Confidence Score: 5/5 **[Low impact]** No blocking defects identified; merge remains subject to PierreOucif's stated requirement to wait for the supporting CLI release. Summary Documents `file-output.combine-certificate-chain` and adds a full-chain example for web servers. - The agent can write a full certificate chain to one file. No blocking defects identified. Merge remains subject to the stated CLI-release prerequisite. ### the check validation blocked The documentation preview could not start because the existing `docs.json` fails schema validation: `logo.href` must be a valid URL, and contextual options `assistant` and `mcp` are unsupported by the cached Mint version. The same fai
… the rest (279 characters)lure occurred with the merge-base page and the updated page. The new table row, full-chain example, and anchor navigation therefore remain unverified in a rendered page. Reviews (1) · Last reviewed commit: "docs(cert-agent): keep server examples t..." · Reviewed by the reviewer
#8485
2026-10-20 23:33
coderabbit
issue-comment
—summary
📝 Walkthrough Adds S3-compatible app connections for supported and configured storage endpoints. Adds admin-managed hostname allowlists and extends PAM session recordings to use AWS or S3-compatible connections. **Changes: S3-Compatible Storage and PAM Recordings** |Layer / File(s)|Summary| |:---|:---| |**Storage hostname configuration** `backend/src/db/migrations/*`, `backend/src/db/schemas/super-admin.ts`, `backend/src/server/routes/v1/admin-router.ts`, `backend/src/server/plugins/serve-ui.ts`, `backend/src/lib/config/env.ts`, `frontend/src/pages/admin/GeneralPage/components/GeneralPageForm.tsx`, `frontend/src/hooks/api/admin/types.ts`, `frontend/index.html`, `docs/documentation/platfo
… the rest (7902 characters)rm/admin-panel/server-admin.mdx`, `docs/self-hosting/configuration/envars.mdx`|Adds persisted and API-managed allowed storage hostnames, an admin form, and HTTPS origins for configured storage hosts in the UI CSP. Adds the internal-IP environment setting.| |**S3-compatible connections** `backend/src/services/app-connection/*`, `backend/src/server/routes/v1/app-connection-routers/*`, `backend/src/lib/aws/s3.ts`, `backend/src/lib/api-docs/constants.ts`, `frontend/src/hooks/api/appConnections/*`, `frontend/src/helpers/appConnections.ts`, `backend/e2e-test/fakes/s3.ts`|Adds S3-compatible connection schemas, endpoint and credential validation, API registration, and frontend connection types and provider resolution. Adds a shared S3 client factory.| |**PAM recording storage** `backend/src/ee/services/pam-session-recording/*`, `backend/src/ee/services/pam-account-template/*`, `backend/src/ee/services/pam-account/*`, `backend/src/ee/services/pam/pam-validators.ts`, `frontend/src/hooks/api/pam/*`, `docs/documentation/platform/pam/sessions/external-storage.mdx`, `backend/src/ee/services/pam/CLAUDE.md`|Resolves AWS or S3-compatible recording access and accepts inherited recording configuration. Recording deletion uses batches of individual object deletion requests.| |**Connection and recording forms** `frontend/src/pages/organization/AppConnections/AppConnectionsPage/components/AppConnectionForm/*`, `frontend/src/pages/pam/PamAccountsPage/components/*`, `frontend/src/pages/pam/PamTemplatesPage/components/TemplateDetailSheet.tsx`|Adds S3-compatible connection creation and editing forms. PAM account and template forms can select AWS or S3-compatible recording connections and omit region settings for S3-compatible connections.| **Priority:** ➖ Normal **Estimated code review effort:** 4 (Complex) | ~60 minutes **Change:** Feature ### Sequence Diagram(s) ```mermaid sequenceDiagram participant AdminForm as GeneralPageForm participant AdminAPI as admin-router participant ConnectionForm as S3CompatibleConnectionForm participant ConnectionAPI as app-connection-router participant CredentialValidation as validateS3CompatibleConnectionCredentials participant PAM as PAM recording services participant Storage as S3 storage AdminForm->>AdminAPI: save allowedStorageHostnames AdminAPI->>AdminAPI: validate and store hostname configuration ConnectionForm->>ConnectionAPI: create or update S3-compatible connection ConnectionAPI->>CredentialValidation: validate endpoint and credentials CredentialValidation->>Storage: call ListBuckets PAM->>PAM: resolve recording connection and configuration PAM->>Storage: send recording storage requests ``` Merge Risk: 🔵 Low · up to b528b Switching an account from AWS to S3-compatible recording storage can make a save fail when the bucket is cleared to inherit the template setting. The issue is narrow and has a straightforward fix. Security architecture risk: 🟡 Moderate · up to b528b Custom storage destinations expand where recording data and storage requests can travel. Endpoint restrictions, access checks, and playback integrity controls provide meaningful protection, but upload-client behavior and failure recovery remain incompletely established. No newly introduced exploitable bypass was established. **Retained concerns** No architecture-level concerns identified. Security review details **Security Blast Radius** - _inferred_ — The new destination policy is instance-wide, while recording connection selection remains organization-bound and requires Connect permission on an organization-scoped connection. An approved custom destination can serve multiple authorized organizations; enabling private-IP access expands reachable backend network scope without itself granting cross-organization recording access. **Trust Boundaries and Controls** - _observed_ — Backend DNS pinning protects backend storage requests, not clients consuming presigned URLs. Presigning does not invoke the configured agent, and browser playback makes its own fetch. The external uploader’s network enforcement remains a material coverage gap, not a verified SSRF finding. **Resilience and Maintainability Implications** - _observed_ — External upload and metadata persistence remain separate operations. Duplicate metadata inserts are ignored, and insertion does not verify the stored object. These behaviors predate this PR; playback integrity checks detect corruption but do not establish upload completion or recover orphaned objects. The inspected deletion implementation supports paginated, repeatable object deletes, but its production lifecycle owner remains unestablished. **Hardening Proposals** - _proposed_ — Establish the actual upload consumer’s DNS, redirect, TLS, retry, and completion guarantees before treating backend endpoint validation as an end-to-end network control. Validate SDK redirect behavior separately. - _proposed_ — For the inherited recording lifecycle, consider durable storage provenance, explicit upload-completion reconciliation, and an identified cleanup owner so destination changes and interrupted writes have defined recovery behavior. Pre-merge checks: 4 · 1 Failed checks: 1 warning | Check name | Status | Explanation | Resolution | | :---------------- | :----------------------------------------------------------------------------------------------: | :-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | :--------------------------------------------------------------------------------- | | Docstring Coverage | | Docstring coverage is 0.00% which is insufficient. The required threshold is 80.00%. Docstring coverage is scoped to functions touched by this diff. Analyzed 2 functions across 42 files. (5 skipped: 5… | Write docstrings for the functions missing them to satisfy the coverage threshold. | Passed checks: 4 passed | Check name | Status | Explanation | | :------------------------ | :--------------------------------------------------------------------------------------------: | :-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | Title check | | The title clearly and concisely describes the main change: adding S3-compatible storage for PAM session recordings. It follows the conventional commit format. | | Description check | | The description includes relevant context, a UI screenshot, the feature classification, checklist items, and an autogenerated implementation summary. The Steps to verify section is empty, and some con… | | Linked Issues check | | Check skipped because no linked issues were found for this pull request. | | Out of Scope Changes check | | Check skipped because no linked issues were found for this pull request. | Full details: Description check **Explanation** The description includes relevant context, a UI screenshot, the feature classification, checklist items, and an autogenerated implementation summary. The Steps to verify section is empty, and some conditional checklist items remain unchecked, but the description is mostly complete and directly related to the pull request. Full details: Docstring Coverage **Explanation** Docstring coverage is 0.00% which is insufficient. The required threshold is 80.00%. Docstring coverage is scoped to functions touched by this diff. Analyzed 2 functions across 42 files. (5 skipped: 5 unsupported.) **Resolution** Write docstrings for the functions missing them to satisfy the coverage threshold. ✨ Finishing Touches: 💡 1 📝 Generate docstrings: 💡 🧪 Generate unit tests (beta) Comment `@the reviewer help` to get the list of available commands.
#8485
2026-10-20 23:42
cubic
inline · P2
backend/src/ee/services/pam-account/pam-account-service.ts:945defect-functional
Key this on changes to the effective recording configuration, not only the patched connection ID: moving an account with an existing connection override to a template with an inherited bucket takes the false branch, so the bucket is not validated and `corsProbeUrl` is null. The UI then skips the bucket CORS check. ```suggestion recordingConnectionId !== undefined || templateId !== undefined || settingsOverrides !== undefined ```
#8485
2026-10-20 23:42
cubic
inline · P2
frontend/src/pages/organization/AppConnections/AppConnectionsPage/components/AppConnectionForm/S3CompatibleConnectionForm.tsx:141defect-functional
The `Method` label targets `id="method"`, but this trigger has no matching ID, so assistive technology cannot associate the label with the selector. Add `id="method"` to the trigger. ```suggestion <SelectTrigger id="method" className="w-full" isError={Boolean(error)}> ```
#8485
2026-10-20 23:42
cubic
inline · P2
frontend/src/helpers/appConnections.ts:710defect-functional
This misses password-only URL userinfo because `username` is empty for `https://:pass@host`; the form accepts it, but backend endpoint validation rejects it. Also reject `url.password` here. ```suggestion url.username || url.password || ```
#8485
2026-10-20 23:42
cubic
inline · P2
frontend/src/helpers/appConnections.ts:681defect-functional
This accepts any `s3.<label>.amazonaws.com` as AWS, so a typo such as `s3.fake-region.amazonaws.com` passes the form and is rejected by backend region validation. Restrict the match to supported AWS regions.
#8485
2026-10-20 23:42
cubic
review-body
—summary
**4 issues found** across 48 files View guided diff | Turn on auto-fix | Re-trigger the reviewer
#8485
2026-10-20 23:42
coderabbit
inline · P2
frontend/src/pages/pam/PamAccountsPage/components/AccountDetailSheet.tsx:638defect-functional
**🗄️ Data Integrity & Integration** | **🟡 Minor** | **⚡ Quick win** **Ignore the hidden region for S3-compatible overrides.** When the selected connection is S3-compatible and the bucket is empty, clear `region` before calculating `hasS3Override`. Otherwise, a stale AWS region can submit `recordingS3Config: { bucket: "" }`. The PATCH schema rejects the empty bucket, so the save fails. 🐛 Proposed fix ```diff --- "a/frontend/src/pages/pam/PamAccountsPage/components/AccountDetailSheet.tsx" +++ "b/frontend/src/pages/pam/PamAccountsPage/components/AccountDetailSheet.tsx" @@ -628,14 +628,14 @@ const onSubmit = (values: OverridesForm) => { const bucket = values.recordingBucket.trim
… the rest (558 characters)(); - const region = values.recordingRegion.trim(); + const region = isS3CompatibleRecording ? "" : values.recordingRegion.trim(); const keyPrefix = values.recordingKeyPrefix.trim(); const hasS3Override = Boolean(bucket) || Boolean(region); const settingsOverrides = hasS3Override ? { recordingS3Config: { bucket, - ...(region && !isS3CompatibleRecording ? { region } : {}), + ...(region ? { region } : {}), ...(keyPrefix ? { keyPrefix } : {}) } } ```
#8485
2026-10-20 23:42
coderabbit
review-body
—summary
**Actionable comments posted: 1**
#8485
2026-10-20 23:43
greptile
issue-comment
—summary
Confidence Score: 4/5 **[High impact]** The PR extends PAM session recording to configured external storage providers\. Not safe to merge until clearing a recording connection validates the effective saved configuration. Findings 1. **Clearing connections breaks recordings** Summary Adds S3-compatible storage connections for PAM recordings, supporting AWS S3, Cloudflare R2, Google Cloud Storage, OCI, and admin-allowed custom hosts. Updates connection forms, recording settings, browser access rules, and setup documentation. One blocking issue remains: clearing an account's recording connection can retain a regionless bucket override while inheriting an AWS connection. The update succee
… the rest (446 characters)ds, but subsequent recording uploads and playback fail. Validate the effective connection and retained settings before merging. - S3-compatible connections support AWS S3 and other object-storage services. - PAM recordings can use compatible storage connections. - Server admins can allow custom storage hosts for session playback. Reviews (1) · Last reviewed commit: "add S3\_COMPATIBLE\_STORAGE\_ALLOW\_INTERNAL..." · Reviewed by the reviewer
#8485
2026-10-20 23:43
greptile
inline · P1
backend/src/ee/services/pam-account/pam-account-service.ts:950defect-functional
**Clearing connections breaks recordings** An account can use an S3-compatible connection with a bucket override that has no region. When its template uses AWS, a PATCH containing only `recordingConnectionId: null` skips this check because `null` is falsy. The update succeeds and preserves the regionless bucket override, but recording access now inherits the AWS connection. Upload and playback requests then fail with “Select the AWS region of the recording bucket.” Validate the effective saved connection and bucket settings whenever the connection changes, including when it is cleared. This recording failure needs to be fixed before merging. Ran code and verified through TREX
#8648
2026-10-20 23:49
cubic
review-body
—summary
**No issues found** across 2 files View guided diff | Turn on auto-fix | Re-trigger the reviewer
#8648
2026-10-20 23:54
coderabbit
issue-comment
—summary
📝 Walkthrough The CLI FAQ and command reference now document that login or logout deletes saved offline secrets. They state that offline access resumes after a later secret fetch. **Changes: Offline secrets documentation** |Layer / File(s)|Summary| |:---|:---| |**Document offline secret lifecycle** `docs/cli/faq.mdx`, `docs/cli/reference.mdx`|The FAQ states that login or logout removes cached secrets and requires a new fetch before offline access resumes. The logout reference clarifies that deletion applies to shared secrets across profiles, and that the next fetch saves them again.| **Priority:** ⬇️ Low **Estimated code review effort:** 1 (Trivial) | ~4 minutes **Change:** Other **S
… the rest (2616 characters)uggested reviewers:** `victorvhs017` Merge Risk: 🔵 Low · up to 7cd62 The FAQ may mislead machine-identity users about offline-secret availability; clarify the conditions before merging. Architecture risk: 🔵 Low · up to 7cd62 The change affects 1 system. **Changed systems:** `docs` **Architecture concerns** No architecture-level concerns identified. Review details **Systems and components** - _observed_ — docs (service) was modified; 2 changed files map to changed impact. **Before / after behavior** - _observed_ — Modified behavior in docs/cli/faq.mdx: Added documentation that login or logout removes cached secrets and requires a new fetch before offline access is available again. - _observed_ — Modified behavior in docs/cli/reference.mdx: The `logout` documentation now states that it deletes shared offline secrets for all profiles, regardless of which profiles remain logged in, and that they are saved again on the next secret fetch. Pre-merge checks: 5 | Check name | Status | Explanation | | :------------------------ | :--------------------------------------------------------------------------------------------: | :-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | Title check | | The title clearly and concisely describes the primary documentation change: documenting that logout deletes the offline secrets backup. It follows the conventional commit format. | | Description check | | The description includes the required context, verification steps, change type, and checklist. It accurately explains the logout behavior and its effect on offline secrets shared across profiles. | | Docstring Coverage | | No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check. Docstring coverage is scoped to functions touched by this diff. Analyzed 0 functions across 0… | | Linked Issues check | | Check skipped because no linked issues were found for this pull request. | | Out of Scope Changes check | | Check skipped because no linked issues were found for this pull request. | Full details: Docstring Coverage **Explanation** No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check. Docstring coverage is scoped to functions touched by this diff. Analyzed 0 functions across 0 files. (2 skipped: 2 unsupported.) ✨ Finishing Touches 🧪 Generate unit tests (beta) Comment `@the reviewer help` to get the list of available commands.
#8648
2026-10-20 23:55
greptile
issue-comment
—summary
Confidence Score: 5/5 **[Low impact]** No blocking defects were identified; merge remains subject to PierreOucif's acknowledged requirement that Infisical/cli#433 be released first. Summary Documents that `infisical logout` deletes secrets saved for offline use. - The CLI docs explain when saved secrets disappear and return. PierreOucif explicitly states that these docs should merge after Infisical/cli#433 is released, because the reference describes the new logout behavior. This remains an acknowledged release dependency, not a reported defect. No blocking issues were found in the documentation changes. Both complete MDX files compiled successfully, and focused browser checks confirmed
… the rest (523 characters) that the new paragraphs are visible. The changes are suitable to merge once the acknowledged CLI release dependency is satisfied. ### the check validation blocked Full Mintlify preview verification was blocked because the pinned Mintlify version rejects existing docs.json settings for logo.href and the contextual options assistant and mcp. Focused rendering succeeded, but full-site behavior was not verified. Reviews (1) · Last reviewed commit: "docs(cli): note that logout deletes the ..." · Reviewed by the reviewer
#8648
2026-10-20 23:58
coderabbit
inline · P2
docs/cli/faq.mdx:66docs
**🎯 Functional Correctness** | **🟡 Minor** | **⚡ Quick win** **Qualify the offline-backup lifecycle.** Machine-identity `infisical login` does not delete the existing backup. Also, only a successful logged-in fetch that requests secret values recreates it. The current sentence can mislead users about offline availability. Suggested fix ```diff --- "a/docs/cli/faq.mdx" +++ "b/docs/cli/faq.mdx" @@ -62,7 +62,7 @@ <Accordion title="Can I fetch secrets with Infisical while I'm offline?"> Yes. If you have previously retrieved secrets for a specific project and environment (such as dev, staging, or prod), the `run`/`secret` command will utilize the saved secrets, even when offline, on subs
… the rest (908 characters)equent fetch attempts. -Running `infisical login` or `infisical logout` deletes the saved secrets, so they're only available offline again after your next fetch. +Running `infisical login` for a user profile or `infisical logout` deletes the saved secrets. The saved secrets become available offline again after a successful logged-in fetch that requests secret values. </Accordion> ``` 📝 Committable suggestion > ‼️ **IMPORTANT** > Carefully review the code before committing. Ensure that it accurately replaces the highlighted code, contains no missing lines, and has no issues with indentation. Thoroughly test & benchmark the code to ensure it meets the requirements. ```suggestion Running `infisical login` for a user profile or `infisical logout` deletes the saved secrets. The saved secrets become available offline again after a successful logged-in fetch that requests secret values. ```
#8648
2026-10-20 23:58
coderabbit
review-body
—summary
**Actionable comments posted: 1**
#8699
2026-10-21 00:00
coderabbit
issue-comment
—summary
No actionable comments were generated in the recent review. 🎉 📝 Walkthrough Release workflows now run a reusable freeze check before changelog generation and release jobs. The check validates the deployment-freeze JSON setting, permits releases when `frozen` is false, and blocks them when the setting is invalid or `frozen` is true. **Changes: Release workflow gating** |Layer / File(s)|Summary| |:---|:---| |**Reusable freeze validation** `.github/workflows/release-freeze-check.yml`|The reusable workflow validates that the input is valid JSON with one object and a boolean `frozen` field. It succeeds when `frozen` is false. It fails for missing or malformed input, or when a freeze is acti
… the rest (4481 characters)ve.| |**Freeze checks in release workflows** `.github/workflows/generate-release-changelog.yml`, `.github/workflows/helm-release-infisical-core.yml`, `.github/workflows/helm-release-infisical-nkp.yml`, `.github/workflows/release-standalone-docker-img-postgres-offical.yml`, `.github/workflows/release_helm_gateway.yaml`|The workflows pass `vars.DEPLOY_FREEZE` to the reusable check. Changelog generation and release jobs now depend on the check. Release jobs retain their test jobs as prerequisites to the check.| **Priority:** ⬇️ Low **Estimated code review effort:** 3 (Moderate) | ~20 minutes **Change:** Other Merge Risk: ⚪ Minimal · up to 81b16 Release workflows now pause for a deployment-freeze check after their existing tests. Valid unfrozen settings allow releases to proceed. Active or invalid settings block releases, as intended. No merge-blocking issue remains. Architecture risk: 🔵 Low · up to 81b16 The changed surface does not map to a changed system, dependency edge, entrypoint, or external dependency. **Changed systems:** None identified. **Architecture concerns** No architecture-level concerns identified. Review details **Before / after behavior** - _observed_ — Modified behavior in .github/workflows/generate-release-changelog.yml: Adds a release freeze-check job using `vars.DEPLOY_FREEZE` and makes `generate_changelog` wait for its completion via `needs`. - _observed_ — Modified behavior in .github/workflows/helm-release-infisical-core.yml: Added a `freeze-check` job that runs after `test-helm`, invokes the release-freeze check workflow, and supplies `vars.DEPLOY_FREEZE` as its `freeze` input. - _observed_ — Modified behavior in .github/workflows/helm-release-infisical-core.yml: Changed the `release` job dependency from `test-helm` to `freeze-check`. - _observed_ — Modified behavior in .github/workflows/helm-release-infisical-nkp.yml: Added the `freeze-check` job, requiring `test-helm` and passing `DEPLOY_FREEZE` to the reusable release-freeze workflow. Changed `release` to depend on `freeze-check` instead of directly on `test-helm`. Pre-merge checks: 4 · 1 Failed checks: 1 warning | Check name | Status | Explanation | Resolution | | :--------------- | :----------------------------------------------------------------------------------------------: | :-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | :---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | Description check | | The description explains the change and classifies it as a chore, but the Steps to verify section is empty and the checklist is entirely unchecked. No test results are provided. | Add concrete verification steps and test results. Complete the checklist by confirming the conventional title, local testing status, documentation status, CLAUDE.md status, and contributing guide review. | Passed checks: 4 passed | Check name | Status | Explanation | | :------------------------ | :--------------------------------------------------------------------------------------------: | :-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | Title check | | The title clearly and concisely describes the main change: validating the freeze flag before release workflows. | | Docstring Coverage | | No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check. Docstring coverage is scoped to functions touched by this diff. Analyzed 0 functions across 0… | | Linked Issues check | | Check skipped because no linked issues were found for this pull request. | | Out of Scope Changes check | | Check skipped because no linked issues were found for this pull request. | Full details: Docstring Coverage **Explanation** No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check. Docstring coverage is scoped to functions touched by this diff. Analyzed 0 functions across 0 files. (6 skipped: 6 unsupported.) ✨ Finishing Touches 🧪 Generate unit tests (beta) Comment `@the reviewer help` to get the list of available commands.
#8699
2026-10-21 00:02
cubic
review-body
—summary
**No issues found** across 6 files View guided diff | Turn on auto-fix | Re-trigger the reviewer
#8699
2026-10-21 00:06
greptile
issue-comment
—summary
Confidence Score: 5/5 **[High impact]** Safe to merge. The release-freeze check blocks frozen or invalid inputs while preserving existing test prerequisites. Summary Adds a shared release-freeze check to Docker, Helm, and changelog workflows. - Release workflows check the freeze setting before running release work. No actionable issues were found. The change is safe to merge based on the executed shell checks and local dependency checks. A hosted release was not executed. Reviews (1) · Last reviewed commit: "move to before the actual release step" · Reviewed by the reviewer
#8710
2026-10-21 00:08
cubic
review-body
—summary
**No issues found** across 2 files View guided diff | Turn on auto-fix | Re-trigger the reviewer
#8710
2026-10-21 00:08
coderabbit
issue-comment
—summary
No actionable comments were generated in the recent review. 🎉 📝 Walkthrough The project-template eligibility list and organization settings links no longer include PAM. The settings list retains KMS. **Changes: Project-template scope** |Layer / File(s)|Summary| |:---|:---| |**Limit project-template products** `frontend/src/helpers/project.ts`, `frontend/src/pages/organization/SettingsPage/components/OrgProductSettingsTab/OrgProductSettingsTab.tsx`|The project helper and organization settings list no longer include PAM for project templates. The settings list retains KMS.| **Priority:** ⬇️ Low **Estimated code review effort:** 1 (Trivial) | ~5 minutes **Change:** Bug fix **Suggested
… the rest (3297 characters) reviewers:** `mathnogueira` Merge Risk: ⚪ Minimal · up to 7550e The change correctly removes PAM project-template support without affecting KMS. Pre-merge checks: 4 · 1 Failed checks: 1 warning | Check name | Status | Explanation | Resolution | | :--------------- | :----------------------------------------------------------------------------------------------: | :-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | :------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ | | Description check | | The description identifies the change through the generated summary and marks the PR as a fix, but the required Context and Steps to verify the change sections are empty. The screenshots are also plac… | Add a concise Context section that explains the previous and new behavior. Add reproducible verification steps. Move the images under the Screenshots section or remove that section if not needed. Complete the applicable checklist items, inc… | Passed checks: 4 passed | Check name | Status | Explanation | | :------------------------ | :--------------------------------------------------------------------------------------------: | :-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | Title check | | The title clearly and concisely describes the main change: removing PAM from the project templates list. | | Docstring Coverage | | No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check. Docstring coverage is scoped to functions touched by this diff. Analyzed 0 functions across 2… | | Linked Issues check | | Check skipped because no linked issues were found for this pull request. | | Out of Scope Changes check | | Check skipped because no linked issues were found for this pull request. | Full details: Docstring Coverage **Explanation** No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check. Docstring coverage is scoped to functions touched by this diff. Analyzed 0 functions across 2 files. Full details: Description check **Explanation** The description identifies the change through the generated summary and marks the PR as a fix, but the required Context and Steps to verify the change sections are empty. The screenshots are also placed before the Screenshots heading, and the checklist remains largely unchecked. **Resolution** Add a concise Context section that explains the previous and new behavior. Add reproducible verification steps. Move the images under the Screenshots section or remove that section if not needed. Complete the applicable checklist items, including local testing and the title-format confirmation. ✨ Finishing Touches 📝 Generate docstrings 🧪 Generate unit tests (beta) Comment `@the reviewer help` to get the list of available commands.
#8710
2026-10-21 00:14
greptile
issue-comment
—summary
Confidence Score: 5/5 **[Low impact]** Safe to merge within the reviewed scope; the intended PAM removal preserves supported template settings. Summary Removes PAM from project-template settings. - PAM no longer appears in shared project-template settings. No actionable issues found. Safe to merge within the reviewed scope. Reviews (1) · Last reviewed commit: "remove PAM from secret template in org s..." · Reviewed by the reviewer
#8637
2026-10-21 00:16
coderabbit
issue-comment
—summary
📝 Walkthrough The change adds AWS KMS-backed signing keys for internal certificate authorities. It introduces a shared key-source service for AWS KMS and HSM, updates CA storage and lifecycle operations, and adds licensing, API, interface, and documentation support. **Changes: External CA key sources** |Layer / File(s)|Summary| |:---|:---| |**Key-source contracts and persistence** `backend/src/db/migrations/*`, `backend/src/db/schemas/certificate-authority-secret.ts`, `backend/src/services/certificate-authority/internal/*schemas.ts`, `backend/src/services/certificate-authority/ca-key-source/*types.ts`, `backend/src/services/signer/signer-enums.ts`, `backend/src/ee/services/license/*`, `b
… the rest (10516 characters)ackend/src/services/license-client/feature-mapping.ts`, `backend/src/lib/api-docs/constants.ts`, `frontend/src/hooks/api/*`|The CA secret storage and configuration contracts now include external key identifiers, public-key data, AWS connection IDs, and regions. The key-source enum and license feature mapping include AWS KMS.| |**External provider signing** `backend/src/services/certificate-authority/ca-key-source/*`, `backend/src/services/certificate-authority/ca-signer.ts`, `backend/src/services/certificate-authority/ca-key-source/aws-kms-ca-key-source.test.ts`, `backend/src/services/certificate/certificate-fns.ts`|A shared service dispatches key operations to HSM or AWS KMS providers. The AWS KMS provider creates and describes keys, signs CA artifacts, and prepares key release. External signer helpers reuse ephemeral keys; tests cover AWS KMS signing, algorithm handling, permissions, and release behavior.| |**CA creation, signing, and key release** `backend/src/services/certificate-authority/internal/internal-certificate-authority-service.ts`, `backend/src/services/certificate-authority/certificate-authority-{fns,service,secret-dal}.ts`, `backend/src/services/project/project-cleanup-queue.ts`, `backend/src/services/app-connection/app-connection-service.ts`, `backend/src/services/cert-manager-export/cert-manager-export-service.ts`, `backend/CLAUDE.md`|CA workflows now create and use external keys through the key-source service. CA and project deletion prepare and execute key releases. Project lookup exposes externally keyed CA secrets; export rejects CAs whose external connections cannot be recreated.| |**CA service and route integration** `backend/src/server/routes/*`, `backend/src/services/certificate-authority/certificate-authority-queue.ts`, `backend/src/services/certificate-profile/*`, `backend/src/services/certificate-v3/*`, `backend/src/services/certificate/certificate-service.ts`, `backend/src/services/pki-application/*`, `backend/src/ee/services/certificate-authority-ocsp/*`, `backend/src/ee/services/pki-scep/pki-scep-fns.ts`|Certificate, SCEP, OCSP, queue, and route dependencies now use the key-source service. Internal CA CSR signing resolves provider-required signature algorithms. CA audit events include key-source metadata.| |**CA setup and key-source display** `frontend/src/pages/cert-manager/CertificateAuthoritiesPage/components/CreateCaWizard/*`, `frontend/src/pages/cert-manager/CertAuthDetailsByIDPage/components/CaDetailsSection.tsx`, `frontend/src/pages/cert-manager/CertificateAuthoritiesPage/components/CaSection.tsx`, `docs/documentation/platform/pki/ca/private-ca.mdx`, `.infisicalignore`|The CA wizard supports AWS connection and region selection, license gating, and source-specific algorithms. CA details and deletion confirmation display AWS KMS key information. The documentation describes AWS KMS setup and deletion timing.| **Priority:** ➖ Normal **Estimated code review effort:** 4 (Complex) | ~60 minutes **Change:** Feature ### Sequence Diagram(s) ```mermaid sequenceDiagram participant User participant CreateCaWizard participant InternalCertificateAuthorityService participant CaKeySourceService participant AWSKMS participant CertificateAuthoritySecretDAL User->>CreateCaWizard: Select AWS KMS, connection, and region CreateCaWizard->>InternalCertificateAuthorityService: Submit CA configuration InternalCertificateAuthorityService->>CaKeySourceService: Create external CA key CaKeySourceService->>AWSKMS: Create key and retrieve public key AWSKMS-->>CaKeySourceService: Key reference and public key CaKeySourceService-->>InternalCertificateAuthorityService: Key details and signer InternalCertificateAuthorityService->>CertificateAuthoritySecretDAL: Persist external key reference ``` **Suggested reviewers:** `carlosmonastyrski`, `sheensantoscapadngan`, `saifsmailbox98` Merge Risk: 🔵 Low · up to 407d6 This change adds AWS KMS-backed signing keys for internal CAs and moves HSM handling onto a shared key-source layer. No correctness or security defect was identified. One efficiency issue remains: each AWS KMS signing operation rebuilds its AWS client and may refresh credentials first. That can add latency, STS throttling risk and AWS cost on high-volume issuance or OCSP traffic. The change is mergeable, with that follow-up recommended. Security architecture risk: 🟡 Moderate · up to 407d6 AWS-backed signing preserves important setup authorization and key-selection controls. However, failed or interrupted cleanup can remove a certificate authority or project without retiring its AWS signing key. The change also requires coordinated schema rollback and appropriately restricted AWS permissions. **Retained concerns** - **Medium · security · inferred:** AWS key retirement is not durably coupled to CA ownership. CA deletion and project hard deletion remove owning database records before executing in-memory release callbacks, while release preparation and execution errors are suppressed. An AWS credential or permission failure, provider failure, or interruption can therefore leave a signing key unreleased after deletion succeeds, without the owning record needed for normal retry. Creation has a related orphan window between AWS allocation and database persistence. Any subsequent signing still requires AWS authority; this is a lifecycle-security concern, not an unauthenticated signing bypass. Security review details **Security Blast Radius** - _inferred_ — The retained lifecycle concern affects the AWS key of an individual CA, or the collected external CA keys during project hard deletion. A surviving key still requires AWS signing authority to use. Application operations target stored key ARNs, but maximum authority of the underlying AWS principal cannot be bounded without deployed IAM policies. **Security Findings and Attack Paths** - _inferred_ — An authorized deletion can complete despite failed AWS retirement because both preparation and release errors are suppressed. Interruption after database deletion produces a similar recovery gap. This can retain signing capability outside the application's CA inventory; it does not demonstrate that an unprivileged caller gains signing permission. **Trust Boundaries and Controls** - _observed_ — CA creation checks project Create permission and provider usage authorization before allocating a key. Connection authorization checks organization ownership, AWS type, Connect permission, and project affinity. Generic CA deletion checks Delete permission and profile dependencies. Runtime signing delegates through a persisted server-created key reference rather than accepting a fresh caller-selected ARN. **Resilience and Maintainability Implications** - _observed_ — Normal failure containment includes duplicate-name prechecking before allocation, transactional CA-secret persistence, attempted cleanup when public-key retrieval or later creation fails, and reading the CA secret from the primary database before deletion. These controls reduce ordinary failure windows but do not durably recover lost or failed release callbacks. **Hardening Proposals** - _proposed_ — Persist key-retirement intent and ownership information before deleting the CA or project, retaining the key ARN and credential reference until retirement is confirmed. Use retryable, idempotent recovery with explicit failed and pending-deletion states, and make allocation recoverable across interruption before CA persistence. - _proposed_ — Validate deployed role trust and KMS policies against the intended organization, project, key, and operation scope. Consider durable ownership tags for reconciliation. These are deployment assurances and hardening proposals, not findings that existing AWS permissions are excessive. Pre-merge checks: 4 · 1 Failed checks: 1 inconclusive | Check name | Status | Explanation | Resolution | | :---------------- | :--------------------------------------------------------------------------------------------------------: | :-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | :--------------------------------------------------------------------------------- | | Docstring Coverage | | Docstring coverage is 0.00% which is insufficient. The required threshold is 80.00%. Docstring coverage is scoped to functions touched by this diff. Analyzed 2 functions across 50 files. (6 skipped: 3… | Write docstrings for the functions missing them to satisfy the coverage threshold. | Passed checks: 4 passed | Check name | Status | Explanation | | :------------------------ | :--------------------------------------------------------------------------------------------: | :-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | Title check | | The title clearly and concisely describes the main change: adding AWS KMS-backed signing keys for internal CAs. It follows the conventional commit format. | | Description check | | The description provides clear context, identifies the feature and documentation changes, and completes the checklist. The Steps to verify section is empty, and the Screenshots placeholder remains des… | | Linked Issues check | | Check skipped because no linked issues were found for this pull request. | | Out of Scope Changes check | | Check skipped because no linked issues were found for this pull request. | Full details: Description check **Explanation** The description provides clear context, identifies the feature and documentation changes, and completes the checklist. The Steps to verify section is empty, and the Screenshots placeholder remains despite UI changes, but the description is otherwise mostly complete. Full details: Docstring Coverage **Explanation** Docstring coverage is 0.00% which is insufficient. The required threshold is 80.00%. Docstring coverage is scoped to functions touched by this diff. Analyzed 2 functions across 50 files. (6 skipped: 3 unsupported, 3 over the file limit.) **Resolution** Write docstrings for the functions missing them to satisfy the coverage threshold. ✨ Finishing Touches: 💡 1 📝 Generate docstrings: 💡 🧪 Generate unit tests (beta) Comment `@the reviewer help` to get the list of available commands.
#8637
2026-10-21 00:24
cubic
inline · P2
backend/CLAUDE.md:1030docs
`pkiAwsKmsSigning` is enforced only by the AWS KMS provider; the HSM provider's `createKey` calls `generateKeyPair` without an entitlement check. Describe the check as provider-specific so future providers do not inherit the AWS-only gate. ```suggestion `createKey` (provider-specific entitlement checks and key creation; returns the columns to store) and `buildSigner` (returns a `TCaSigner`, usually through ```
#8637
2026-10-21 00:24
cubic
inline · P2
backend/src/services/certificate-authority/ca-key-source/ca-key-source-service.ts:106defect-data
AWS credential, network, or API failures are swallowed here, so CA deletion can succeed while its KMS key remains active with no retry path. Persist a cleanup retry or otherwise retain the key reference until deletion is scheduled.
#8637
2026-10-21 00:24
cubic
inline · P2
docs/documentation/platform/pki/ca/private-ca.mdx:288docs
The new AWS KMS algorithms contradict the page’s global “What key algorithms are supported?” answer, which still lists only RSA 2048/4096 and ECDSA P-256/P-384. Update the FAQ and distinguish AWS KMS support from the other key sources so readers see the correct options.
#8637
2026-10-21 00:24
cubic
inline · P2
backend/src/db/migrations/20261012152847_ca-secret-external-key.ts:57defect-data
Rolling back after an AWS KMS CA has been created fails here because the replacement check rejects every `keySource = 'aws-kms'` row. Handle or explicitly guard these rows before attempting the legacy constraint so rollback does not fail at constraint creation.
#8637
2026-10-21 00:24
cubic
inline · P2
backend/src/services/certificate-authority/certificate-authority-secret-dal.ts:14defect-data
This query reads from a replica, so replication lag can hide an external CA secret from the one-time cleanup; the project cascade then removes its reference without scheduling KMS deletion. Default to the primary. ```suggestion (tx || db.primaryNode())(TableName.CertificateAuthoritySecret) ```
#8637
2026-10-21 00:24
cubic
review-body
—summary
**7 issues found** across 56 files View guided diff | Turn on auto-fix | Re-trigger the reviewer
#8637
2026-10-21 00:24
cubic
inline · P2
backend/src/services/certificate-authority/internal/internal-certificate-authority-schemas.ts:94defect-security
The deprecated create endpoint still accepts caller-supplied read-only key metadata through this shared schema. Apply the read-only refinement to its create schema too.
#8637
2026-10-21 00:24
cubic
inline · P2
backend/src/services/certificate-authority/ca-key-source/aws-kms-ca-key-source.test.ts:240tests
This skips the ECDSA KMS regression test whenever OpenSSL lacks ML-DSA support. Use Node crypto for the ECDSA fake signer or gate this test on ECDSA-specific availability so CI still exercises the path without PQC.
#8637
2026-10-21 00:24
coderabbit
inline · P2
backend/src/services/certificate-authority/ca-key-source/aws-kms-ca-key-source.ts:188defect-performance
**🚀 Performance & Scalability** | **🟡 Minor** | **⚡ Quick win** **Cache the KMS client across signers; do not rebuild it on every `buildSigner` call.** `client` lives inside `buildSigner`. Each `getCaSigner` call therefore runs `$getKmsClient` again for an AWS KMS CA. That function reads the app connection, decrypts it, and calls `getAwsConnectionConfig`. For assume-role connections, that is an STS `AssumeRole` round trip before every KMS `Sign`. Every issuance, CRL rebuild and OCSP miss pays this cost. The OCSP responder is unauthenticated, and nonced requests are never cached. An anonymous caller can therefore drive one decrypt, one STS call and one KMS `Sign` per request. On self-hos
… the rest (1590 characters)ted instances, only the signing concurrency cap limits this. The cost is latency, STS throttling risk and KMS charges. Keep a provider-level cache keyed by `appConnectionId` and region. Give it a TTL shorter than the 15-minute assumed-role session. Drop the entry when a send fails with a credential error. ♻️ Proposed change ```diff --- "a/backend/src/services/certificate-authority/ca-key-source/aws-kms-ca-key-source.ts" +++ "b/backend/src/services/certificate-authority/ca-key-source/aws-kms-ca-key-source.ts" @@ -183,9 +183,8 @@ }); } - let client: KMSClient | undefined; const sign = async (message: Buffer, messageType: MessageType, signingAlgorithm: SigningAlgorithmSpec) => { - client ??= await $getKmsClient(appConnectionId, parseAwsKmsKeyArnRegion(keyArn)); + const client = await $getCachedKmsClient(appConnectionId, parseAwsKmsKeyArnRegion(keyArn)); try { const { Signature } = await client.send( new SignCommand({ ``` ```ts // at factory scope const KMS_CLIENT_TTL_MS = 10 * 60 * 1000; const kmsClients = new Map<string, { client: Promise<KMSClient>; expiresAt: number }>(); const $getCachedKmsClient = (appConnectionId: string, region: string) => { const key = `${appConnectionId}:${region}`; const hit = kmsClients.get(key); if (hit && hit.expiresAt > Date.now()) return hit.client; const client = $getKmsClient(appConnectionId, region); kmsClients.set(key, { client, expiresAt: Date.now() + KMS_CLIENT_TTL_MS }); client.catch(() => kmsClients.delete(key)); return client; }; ```
#8637
2026-10-21 00:24
coderabbit
review-body
—summary
**Actionable comments posted: 1**
#8637
2026-10-21 00:31
greptile
issue-comment
—summary
Confidence Score: 4/5 **[Medium impact]** The changes reach internal CA setup\, signing\, and key cleanup\. Do not merge until approval and fresh-CSR renewal consistently resolve the certificate signature algorithm for AWS KMS-backed CAs. Findings 1. **CSR approval and renewal fail** Summary This PR adds AWS KMS-backed signing keys for internal CAs. - Internal CAs can create signing keys in AWS KMS and use them through an AWS connection. - CA signing workflows send signature requests through the selected key source. - External CA keys use shared storage and are released when their CA is deleted. - The CA wizard lets administrators choose AWS KMS and see the key details. **Fix before
… the rest (431 characters)merging:** CSR approval and fresh-CSR renewal still need the signature correction used by initial signing. An AWS KMS P-384 CA accepts a request under a SHA384-only policy, but approval rejects the same request under the unchanged policy. Fresh-CSR renewal also fails, either during policy validation or at the KMS provider's digest check. Reviews (1) · Last reviewed commit: "Add alias to AWS KMS keys" · Reviewed by the reviewer
#8637
2026-10-21 00:31
greptile
inline · P1
backend/src/services/certificate-v3/certificate-v3-service.ts:1435defect-functional
**CSR approval and renewal fail** This correction runs only during initial CSR signing. For an AWS KMS P-384 CA, submission resolves a SHA-256 CSR to `ECDSA-SHA384`, so a policy permitting only `ECDSA-SHA384` accepts the request. Approval then reconstructs `ECDSA-SHA256` from the original CSR and rejects the same request under the unchanged policy. Fresh-CSR renewal also skips this correction: the SHA384-only policy rejects it, and even when the policy permits SHA256, signing fails because the KMS P-384 provider requires SHA384. Apply the same CA-aware resolution before approval and renewal validation, and pass the resolved algorithm through to signing. This must be fixed before merging be
… the rest (1443 characters)cause accepted requests cannot be approved and existing certificates cannot be renewed through this path. Artifacts **Before control responses with the base submission service** - Ran the exact base submission service against the same downstream contracts and captured HTTP responses, commands, exit codes, source excerpts, and the complete executed harness; submission rejects the CSR before approval. **Head responses showing accepted submission and failing approval and renewal** - Ran unmodified head services and captured HTTP responses, call arguments, commands, exit codes, source excerpts, and the complete executed harness; approval and renewal fail without downstream algorithm normalization. **Executed CSR approval and renewal regression checks** - Generated real CSR and CA fixtures and exercised real service functions through local HTTP adapters with mocked infrastructure; both final runs passed assertions reproducing the supplied failure. **Executed configuration for the base-submission and head controls** - Loaded the exact base submission source only for the before control while retaining identical downstream services; the comparison isolates submission normalization. **Executed command and evidence capture script** - Executed both controls and saved their actual combined output with reproducible metadata and source; both captured Vitest processes exited successfully. Ran code and verified through TREX
538 of 538 defects
DefectSeverityWhat the user meetsThe testBlamed PRsFound by a bot
Checking an emailed code reads every user's pending codes to find one
s4-recovery-token-lookup-scans-token-table-1
account-recoveryperformanceloss: time
medium

When Anyone checks an emailed code — a recovery link, a password setup link, an email sign-in code.

Got The product looks through every pending code of every user to find yours, so the check gets slower as more people use the product.

Want Your code is found directly, at the same cost however many users there are.

steps
  1. Request an account recovery email for an account
  2. Open the recovery link from the email so the code is checked
  3. Look at the indexes on the emailed-token table (auth_tokens) in the database the product's migrations built
s4-recovery-token-lookup-scans-token-table
e2e/specs/account-recovery/s4-recovery-token-lookup-scans-token-table.spec.ts:10
lines
infisical/backend/src/db/migrations/20231129072939_auth-token.ts:8
infisical/backend/src/services/auth-token/auth-token-service.ts:206
infisical/backend/src/services/auth-token/auth-token-service.ts:232
infisical/backend/src/services/auth-token/auth-token-service.ts:122
infisical/backend/src/services/account-recovery/account-recovery-service.ts:182
before the PRs testedno
Two password changes at once are both reported as saved, but only one password works
s4-concurrent-password-changes-both-saved-1
account-recoverybehaviourloss: decision
medium

When You change your password twice at the same moment, from two tabs, with the same current password.

Got Both changes say the password was changed. One of the two new passwords does not sign you in.

Want Only one change goes through; the other says the current password is wrong.

steps
  1. Create an account with password P0 and sign in.
  2. As a control, on another account: change P0 to P1, sign in with P1, then try to change from P0 to P2 — it is refused with "Incorrect current password.".
  3. On the first account, send two password changes with the same current password P0 at the same moment — one to P1, one to P2 (two tabs, or two API calls) — with the save of the user row held until both have read the account.
  4. Look at both answers, then sign in with P1 and with P2.
s4-concurrent-password-changes-both-saved
e2e/specs/account-recovery/s4-concurrent-password-changes-both-saved.spec.ts:27
lines
infisical/backend/src/services/auth/auth-password-service.ts:60
infisical/backend/src/services/auth/auth-password-service.ts:82
infisical/backend/src/services/auth/auth-password-service.ts:90
infisical/backend/src/server/routes/v2/password-router.ts:53
before the PRs testedno
The choose-organization screen spins forever when the account fails to load
s4-select-org-user-error-loader-forever-1
behaviourloss: time
medium

When

Got

Want

steps
  1. Sign in with an account that belongs to an organization
  2. Have the service that returns the signed-in user answer with a server error
  3. Open the choose-organization screen (/login/select-organization?force=true)
  4. Wait 40 seconds
lines
infisical/frontend/src/pages/auth/SelectOrgPage/SelectOrgPage.tsx:444
infisical/frontend/src/pages/auth/SelectOrgPage/SelectOrgPage.tsx:141
infisical/frontend/src/hooks/api/users/queries.tsx:41
before the PRs testedno
Choosing an organization after sign-in gets slower as the whole instance grows
s10-org-picker-scans-all-memberships-1
sign-inperformanceloss: time
medium

When You sign in and reach the screen where you choose an organization.

Got Finding your organizations means looking through every membership of every user in the instance, so the screen slows as the instance grows.

Want Your organizations are found directly, at the same speed however large the instance is.

steps
  1. Sign in as any user and reach the select-organization screen
  2. The screen loads the user's organizations (the accessible-with-sub-orgs list)
  3. Read the database's index list for the memberships table
s10 memberships of one user are found through an index on the user
e2e/specs/sign-in/s10-org-picker-scans-all-memberships.spec.ts:6
lines
infisical/backend/src/services/org/org-dal.ts:258
infisical/backend/src/services/org/org-dal.ts:263
infisical/backend/src/services/org/org-dal.ts:425
infisical/backend/src/db/migrations/20251005152640_simplify-membership.ts:47
infisical/backend/src/db/migrations/20260317200000_membership-unique-constraints.ts:169
before the PRs testedno
Account recovery claims email is not set up when the server check fails
s1-status-check-failure-reads-as-email-not-configured-1
behaviourloss: time
medium

When

Got

Want

steps
  1. Sign out and open the account recovery page while the server status check fails with a server error.
  2. Enter an email address in the recovery form.
  3. Press Continue.
lines
infisical/frontend/src/pages/auth/AccountRecoveryEmailPage/AccountRecoveryEmailPage.tsx:49
infisical/frontend/src/pages/auth/AccountRecoveryEmailPage/AccountRecoveryEmailPage.tsx:20
infisical/frontend/src/hooks/api/serverDetails/queries.tsx:17
before the PRs testedno
A server failure on a valid recovery link says the email was not verified
s1-confirm-email-failure-reported-as-unverified-1
behaviourloss: time
medium

When

Got

Want

steps
  1. Sign out and request an account recovery email for an email account.
  2. Open the recovery link from the email; the Confirm your email step shows.
  3. Make the verification request fail on the server (here: a 500 injected on the verify-email request).
  4. Press Confirm Email.
lines
infisical/frontend/src/pages/auth/AccountRecoveryResetPage/components/ConfirmEmailStep.tsx:52
infisical/frontend/src/pages/auth/AccountRecoveryResetPage/components/ConfirmEmailStep.tsx:54
infisical/frontend/src/pages/auth/EmailNotVerifiedPage/EmailNotVerifiedPage.tsx:17
before the PRs testedno
A failed load of the SSO settings says no identity provider is connected
s1-failed-provider-load-reads-as-none-1
behaviourloss: time
medium

When

Got

Want

steps
  1. As an organisation admin with OIDC single sign-on configured, open Organization Settings > SSO & Provisioning and see the OIDC card with its Enable OIDC toggle.
  2. Make the OIDC, SAML and LDAP settings requests fail with a server error.
  3. Reload the page.
lines
infisical/frontend/src/hooks/api/oidcConfig/queries.tsx:17
infisical/frontend/src/hooks/api/ssoConfig/queries.tsx:22
infisical/frontend/src/hooks/api/ldapConfig/queries.tsx:22
infisical/frontend/src/pages/organization/SettingsPage/components/OrgSsoTab/OrgSsoTab.tsx:154
before the PRs testedno
The emailed MFA code stops working while the screen still promises attempts
s11-mfa-code-dead-while-attempts-promised-1
sign-inbehaviourloss: time
medium

When You mistype the emailed sign-in code three times.

Got The screen says you still have 2 tries, but even the right code is now refused, and the next try signs you out as locked.

Want The right code works while tries remain, or the screen says the code expired and offers a new one.

steps
  1. Sign up a new account and turn on enforced MFA for its organization
  2. Sign out, then sign in with email and password to reach the email MFA code prompt
  3. Read the six-digit code from the email in Mailpit
  4. Enter a different six-digit code three times; the screen says 'Invalid code. You have 2 attempt(s) remaining.'
  5. Enter the correct code from the email
s11-mfa-code-dead-while-attempts-promised
e2e/specs/sign-in/s11-mfa-code-dead-while-attempts-promised.spec.ts:7
lines
infisical/frontend/src/components/auth/Mfa.tsx:39
infisical/frontend/src/components/auth/Mfa.tsx:186
infisical/backend/src/services/auth-token/auth-token-service.ts:69
infisical/backend/src/services/auth-token/auth-token-service.ts:224
before the PRs testedno
A server failure during LDAP sign-in is shown as wrong credentials
s18-ldap-server-error-blames-credentials-1
sso-loginbehaviourloss: decision
medium

When You sign in with LDAP while the server has a problem.

Got You are told to double-check your credentials, though they are fine.

Want You are told the sign-in failed on the server and to try again later.

steps
  1. Sign out and open the LDAP login page
  2. Enter organization A's slug, LDAP username john and his password
  3. Have the server fail the LDAP sign-in with an internal error
  4. Press Continue with LDAP
s18-ldap-server-error-blames-credentials
e2e/specs/sso-login/s18-ldap-server-error-blames-credentials.spec.ts:7
lines
infisical/frontend/src/pages/auth/LoginLdapPage/LoginLDAPPage.tsx:67
infisical/frontend/src/pages/auth/LoginLdapPage/LoginLDAPPage.tsx:69
infisical/backend/src/ee/routes/v1/ldap-router.ts:127
before the PRs testedno
Every sign-in from a new browser adds to the user's device list forever
s11-device-list-grows-every-new-browser-1
sign-inperformanceloss: time
medium

When You sign in from a browser or network the product has not seen before.

Got It remembers that device forever; the list keeps growing with every browser update or new network.

Want Only a bounded number of recent devices is kept.

steps
  1. Sign the user in with email and password and pick their organization, from a browser whose user agent was never used before
  2. Repeat 5 times, each with a different user agent, and count the entries of the user's device list
  3. Repeat 50 more times, each with a different user agent, and count again
s11 the device list stays bounded across sign-ins from new browsers
e2e/specs/sign-in/s11-device-list-grows-every-new-browser.spec.ts:27
lines
infisical/backend/src/services/auth/auth-login-service.ts:124
infisical/backend/src/services/auth/auth-login-service.ts:128
infisical/backend/src/services/auth/auth-login-service.ts:1308
infisical/backend/src/services/auth-token/auth-token-service.ts:385
before the PRs testedno
Two tabs refreshing at the same time get signed out as token reuse
s11-concurrent-refresh-revokes-session-1
sign-inbehaviourloss: time
medium

When You have the app open in two tabs and they refresh your session at the same moment.

Got One tab keeps working, but using the other signs you out with a message that your session was revoked for security.

Want Both tabs keep working and you stay signed in.

steps
  1. Sign in so the browser holds a valid session refresh token
  2. From two tabs sharing that cookie, refresh the same token at the same moment
  3. Observe both refreshes succeed and each returns a rotated token
  4. Use each returned token to refresh the session once more
s11-concurrent-refresh-revokes-session
e2e/specs/sign-in/s11-concurrent-refresh-revokes-session.spec.ts:28
lines
infisical/backend/src/services/auth-token/auth-token-service.ts:287
infisical/backend/src/services/auth-token/auth-token-service.ts:296
infisical/backend/src/services/auth-token/auth-token-service.ts:307
infisical/backend/src/services/auth-token/auth-token-service.ts:325
infisical/backend/src/services/auth-token/auth-token-service.ts:328
before the PRs testedno
LDAP group mappings vanish without a word when they fail to load
s9-failed-ldap-group-maps-load-shows-nothing-1
behaviourloss: decision
medium

When

Got

Want

steps
  1. As an organization admin with LDAP configured, save an LDAP group mapping (LDAP group infisical-users to an Infisical group).
  2. Open Organization Settings > SSO & Provisioning and press Configure under LDAP Group Mappings: the mapping is listed.
  3. Reload the page while the group mappings request fails with a server error, and open the LDAP Group Mappings dialog again.
lines
infisical/frontend/src/pages/organization/SettingsPage/components/OrgSsoTab/LDAPGroupMapModal.tsx:84
infisical/frontend/src/pages/organization/SettingsPage/components/OrgSsoTab/LDAPGroupMapModal.tsx:210
infisical/frontend/src/pages/organization/SettingsPage/components/OrgSsoTab/LDAPGroupMapModal.tsx:256
infisical/frontend/src/hooks/api/ldapConfig/queries.tsx:36
before the PRs testedno
A refused OIDC sign-in ends on a raw error page instead of the app
s19-refused-oidc-sign-in-shows-raw-json-1
sso-loginbehaviourloss: time
low

When Your organization's identity provider turns down your OIDC sign-in.

Got You are left on a bare page of code-like text saying Authentication error, outside the app, with no way back.

Want You land on Infisical's sign-in error page with a clear message and a way back to sign in.

steps
  1. Sign out and open the OIDC login page.
  2. Enter a work email at oidc.com (organization A's verified domain with OIDC configured) and press Continue with OIDC.
  3. At the organization's identity provider (Keycloak), have it refuse the sign-in so it sends the browser back to Infisical with an error (the test asks Keycloak for a silent sign-in with no session, which it answers with login_required).
s19-refused-oidc-sign-in-shows-raw-json
e2e/specs/sso-login/s19-refused-oidc-sign-in-shows-raw-json.spec.ts:5
lines
infisical/backend/src/ee/routes/v1/oidc-router.ts:139
infisical/backend/src/ee/routes/v1/oidc-router.ts:176
infisical/backend/src/server/routes/v1/sso-router.ts:568
infisical/frontend/src/routes.ts:479
before the PRs testedno
Two LDAP sign-ins at once give one member two LDAP links
s20-concurrent-ldap-sign-ins-duplicate-link-1
sso-loginbehaviourloss: time
low

When An invited member's first LDAP sign-in is sent twice at the same moment, for example from two tabs.

Got Both go through and the member ends up linked to the organization twice.

Want The member has one LDAP link, whatever the timing.

steps
  1. As an organization admin, verify an email domain, configure LDAP single sign-on and turn it on.
  2. Invite a directory user's email to the organization (they become an invited member).
  3. Send that user's first LDAP sign-in twice at the same moment, both with correct credentials.
  4. Count the user's LDAP links in the organization.
s20-concurrent-ldap-sign-ins-duplicate-link
e2e/specs/sso-login/s20-concurrent-ldap-sign-ins-duplicate-link.spec.ts:90
lines
infisical/backend/src/ee/services/ldap-config/ldap-config-service.ts:501
infisical/backend/src/ee/services/ldap-config/ldap-config-service.ts:590
infisical/backend/src/db/migrations/20260826120000_add-user-alias-lookup-indexes.ts:10
before the PRs testedno
Two admins can leave an org with OIDC sign-in and Google enforcement both on
s21-oidc-and-google-enforcement-both-on-1
sso-loginbehaviourloss: decision
medium

When One admin enforces Google sign-in while another turns on OIDC sign-in at the same moment.

Got Both changes say they saved. The organization ends with Google enforced and OIDC on, a combination the settings refuse when done one at a time.

Want One of the two changes is refused, as it is when they are made one after the other.

steps
  1. Take an organization with an OIDC single sign-on configuration that is off, whose members have signed in with Google at least once.
  2. Control: turn OIDC on, then try to enforce Google sign-in — it is refused ('Cannot enable Google OAuth enforcement while OIDC SSO is configured'); turn OIDC off again.
  3. Admin A enforces Google sign-in; while that change is being saved, admin B turns OIDC single sign-on on.
  4. Both changes report success.
  5. Read the organization: Google sign-in is enforced and OIDC single sign-on is on.
OIDC SSO and Google sign-in enforcement are never both on
e2e/specs/sso-login/s21-oidc-and-google-enforcement-both-on.spec.ts:13
lines
infisical/backend/src/ee/services/oidc/oidc-config-service.ts:644
infisical/backend/src/ee/services/oidc/oidc-config-service.ts:657
infisical/backend/src/ee/services/oidc/oidc-config-service.ts:702
infisical/backend/src/services/org/org-service.ts:595
infisical/backend/src/services/org/org-service.ts:642
infisical/backend/src/services/org/org-service.ts:679
before the PRs testedno
Each OIDC sign-in with group sync loads every group in the organization
s21-oidc-sign-in-reads-every-org-group-1
sso-loginperformanceloss: time
low

When Your organization uses OIDC sign-in with group sync and has many groups.

Got Every sign-in loads all of the organization's groups, with their roles and permissions, to find the few you belong to. The work grows with each group added.

Want Sign-in looks up only the groups your identity provider says you are in.

steps
  1. Create an organization, verify an email domain for it, and turn on OIDC single sign-on against Keycloak with group membership sync enabled.
  2. In Keycloak, put the signing-in user in one group whose name matches one Infisical group of the organization.
  3. Create 5 Infisical groups in the organization (one of them the matching one) and sign in through OIDC as that user, counting the rows of the organization-groups query.
  4. Grow the organization to 50 Infisical groups and sign in again, counting the same query.
an OIDC sign-in with group sync reads only the groups the identity provider names
e2e/specs/sso-login/s21-oidc-sign-in-reads-every-org-group.spec.ts:39
lines
infisical/backend/src/ee/services/oidc/oidc-config-service.ts:455
infisical/backend/src/ee/services/group/group-dal.ts:57
infisical/backend/src/ee/services/oidc/oidc-config-fns.ts:40
before the PRs testedno
Projects look missing when the project list fails to load
s1-failed-project-list-shown-as-no-projects-1
behaviourloss: decision
medium

When

Got

Want

steps
  1. Sign in as a user who belongs to several Secrets Management projects.
  2. Make the project list request fail with a server error.
  3. Open Secrets Management in the organisation, My Projects view.
lines
infisical/frontend/src/pages/organization/ProjectsPage/ProjectTypePage/ProjectTypePage.tsx:426
infisical/frontend/src/pages/organization/ProjectsPage/ProjectTypePage/ProjectTypePage.tsx:600
infisical/frontend/src/pages/organization/ProjectsPage/ProjectTypePage/ProjectTypePage.tsx:726
infisical/frontend/src/hooks/api/reactQuery.tsx:343
before the PRs testedno
The My Projects tab downloads every project you belong to to show 24
s8-my-projects-loads-every-project-1
projectsperformanceloss: time
medium

When You belong to many projects and open the My Projects tab.

Got The page downloads every project you are in, of every product type, with all their environments, then shows 24.

Want The page loads only the 24 projects it shows.

steps
  1. Sign in as a user who belongs to 55 or more Secrets Management projects in one organisation
  2. Open Secrets Management projects; the My Projects tab is selected and shows 24 per page
  3. Watch the project list the page downloads
the My Projects tab loads at most one page of projects
e2e/specs/projects/s8-my-projects-loads-every-project.spec.ts:11
lines
infisical/backend/src/services/project/project-dal.ts:319
infisical/backend/src/server/routes/v1/project-router.ts:324
infisical/frontend/src/pages/organization/ProjectsPage/ProjectTypePage/ProjectTypePage.tsx:426
infisical/frontend/src/pages/organization/ProjectsPage/ProjectTypePage/ProjectTypePage.tsx:486
infisical/frontend/src/hooks/api/projects/queries.tsx:96
before the PRs testedno
Certificate Manager claims no project is selected when its setup fails to load
s8-failed-cert-manager-setup-shown-as-not-configured-1
projectsuiloss: time
medium

When You open Certificate Manager from the Organization Overview and its setup can't be loaded.

Got A dialog says an admin must select the active Certificate Manager project, though one already exists, and no error appears.

Want You are told the setup could not be loaded and can try again.

steps
  1. Sign in as an organisation admin whose organisation has a Certificate Manager project.
  2. Make the Certificate Manager setup request fail (the server answers an error).
  3. Open the Organization Overview.
  4. Click the Certificate Manager tile.
a failed Certificate Manager setup load is not shown as 'Select an active Certificate Manager project'
e2e/specs/projects/s8-failed-cert-manager-setup-shown-as-not-configured.spec.ts:10
lines
infisical/frontend/src/pages/organization/ProjectsPage/components/ProjectCategoryOverview.tsx:128
infisical/frontend/src/pages/organization/ProjectsPage/components/ProjectCategoryOverview.tsx:444
infisical/frontend/src/hooks/api/certManagerInstance/index.ts:32
infisical/frontend/src/components/projects/CertManagerNotConfiguredModal.tsx:45
before the PRs testedno
LDAP sign-in gets slower with each newly mapped group
s20-ldap-login-group-sync-per-group-1
sso-loginperformanceloss: time
low

When Your admin maps your LDAP group to several Infisical groups and you sign in with LDAP.

Got The sign-in adds you to each group separately, so it takes longer the more groups were mapped.

Want Joining many mapped groups costs the sign-in no more than joining one.

steps
  1. As an organization admin with LDAP sign-in enabled and the users' email domain verified, map one LDAP group to several Infisical groups on the SSO page.
  2. As a member of that LDAP group who is not yet in those Infisical groups, sign in with LDAP.
  3. Count the database statements the sign-in makes with one newly mapped group and with five.
lines
infisical/backend/src/ee/services/ldap-config/ldap-config-service.ts:694
infisical/backend/src/ee/services/ldap-config/ldap-config-service.ts:720
infisical/backend/src/ee/services/group/group-fns.ts:188
infisical/backend/src/ee/services/group/group-fns.ts:214
infisical/backend/src/ee/services/group/group-fns.ts:297
before the PRs testedno
Searching All Projects for a name with an underscore says no project matches
s2-project-search-misses-underscore-names-1
behaviourloss: time
medium

When

Got

Want

steps
  1. Sign in as an organisation admin and open Secrets Management.
  2. Click New Project, name it something containing an underscore (e.g. billing_api) and create it.
  3. Go back to the Secrets Management projects list and switch to the All Projects tab.
  4. Type the project's exact name in the search box.
lines
infisical/backend/src/server/routes/v1/project-router.ts:1195
infisical/backend/src/server/routes/v1/project-router.ts:235
infisical/frontend/src/pages/organization/ProjectsPage/ProjectTypePage/ProjectTypePage.tsx:824
infisical/frontend/src/pages/organization/ProjectsPage/ProjectTypePage/ProjectTypePage.tsx:874
infisical/frontend/src/pages/organization/ProjectsPage/ProjectTypePage/ProjectTypePage.tsx:1085
before the PRs testedno
Project search has no page-size cap, so one call returns every project
s8-project-search-accepts-any-limit-1
projectsperformanceloss: time
medium

When A script or tool searches your organisation's projects and asks for a very large page.

Got It gets every matching project in one answer, however many the organisation holds.

Want A page holds at most 100 projects, as the identity and certificate searches do.

steps
  1. Sign in as an organisation admin and get an access token
  2. Create 101 Secrets Management projects whose names share one unique prefix
  3. Call the project search API for that prefix with no limit: 100 projects come back
  4. Call it again for the same prefix with limit 100000
project search refuses or clamps a limit far above 100
e2e/specs/projects/s8-project-search-accepts-any-limit.spec.ts:11
lines
infisical/backend/src/server/routes/v1/project-router.ts:1189
infisical/backend/src/services/project/project-dal.ts:893
infisical/backend/src/server/routes/v1/deprecated-project-router.ts:604
before the PRs testedno
The project list across sub-organisations gets slower with every sub-organisation
s8-sub-org-project-list-reads-membership-per-sub-org-1
projectsperformanceloss: time
medium

When Your organisation has many sub-organisations and you open the project list that spans them.

Got Each extra sub-organisation adds more work to every load of that list, so it slows as your organisation grows.

Want The list takes the same effort to load however many sub-organisations there are.

steps
  1. As an organisation admin on a plan with sub-organisations, create 2 sub-organisations
  2. Load the project list across sub-organisations (the command menu's project list) and count the database statements
  3. Create 18 more sub-organisations
  4. Load the same list again and count the statements
listing projects across sub-organisations reads memberships in a fixed number of statements
e2e/specs/projects/s8-sub-org-project-list-reads-membership-per-sub-org.spec.ts:10
lines
infisical/backend/src/services/project/project-service.ts:958
infisical/backend/src/services/project/project-service.ts:960
infisical/backend/src/services/org/org-dal.ts:933
infisical/frontend/src/hooks/api/projects/queries.tsx:151
before the PRs testedno
A project is deleted even though its delete protection was just saved as on
s9-delete-protection-saved-during-delete-lost-1
projectsbehaviourloss: work
medium

When One admin turns on delete protection for a project while another deletes it at the same moment.

Got Both are told they succeeded, and the project is deleted anyway.

Want The project stays protected and the delete is refused, or the protection save says the project is gone.

steps
  1. Create a secrets project with delete protection off.
  2. As one admin, turn delete protection on in the project's settings and save.
  3. At the same moment, as another admin, delete the project from its settings.
  4. Reload the project.
a project whose delete protection is saved during its delete is kept
e2e/specs/projects/s9-delete-protection-saved-during-delete-lost.spec.ts:11
lines
infisical/backend/src/services/project/project-service.ts:793
infisical/backend/src/services/project/project-service.ts:805
infisical/backend/src/services/project/project-service.ts:861
infisical/backend/src/services/project/project-dal.ts:123
infisical/backend/src/services/project/project-service.ts:1053
before the PRs testedno
Moving an environment down fails after another environment was deleted
s3-move-down-fails-after-environment-deleted-1
behaviourloss: time
low

When

Got

Want

steps
  1. Create a Secrets Management project (it has Development, Staging, Production).
  2. In Settings > Environments, delete Staging (it shows Pending deletion).
  3. Create an environment named qa.
  4. Click Move down on Production.
lines
infisical/frontend/src/pages/secret-manager/SettingsPage/components/EnvironmentSection/EnvironmentTable.tsx:172
infisical/backend/src/services/project-env/project-env-service.ts:219
infisical/backend/src/services/project-env/project-env-dal.ts:36
infisical/backend/src/services/project-env/project-env-dal.ts:160
before the PRs testedno
Two Certificate Manager deletes at once can leave the organisation with none
s10-concurrent-cert-manager-deletes-remove-last-1
projectsbehaviourloss: time
low

When Two Certificate Manager projects are deleted at the same moment and they are the organisation's only two.

Got Both deletes go through and the organisation is left with no Certificate Manager project.

Want One delete goes through and the other is refused because it is the last one.

steps
  1. Create a new organisation; it starts with one Certificate Manager project, none set active
  2. Create a second Certificate Manager project in it
  3. Delete one, then the other: the second is refused as the last Certificate Manager project
  4. Create another Certificate Manager project so there are two again
  5. Delete both at the same moment (two tabs or two admins), with both requests reading the project count before either delete is written
  6. Open the organisation's Certificate Manager setup
two Certificate Manager project deletes at once still keep the organisation's last one
e2e/specs/projects/s10-concurrent-cert-manager-deletes-remove-last.spec.ts:31
lines
infisical/backend/src/services/project/project-service.ts:825
infisical/backend/src/services/project/project-service.ts:829
infisical/backend/src/services/project/project-service.ts:846
infisical/backend/src/services/project/project-service.ts:861
infisical/backend/src/services/project/project-dal.ts:123
before the PRs testedno
Looking up an external KMS by a wrong name reports a server fault, not not-found
s11-external-kms-unknown-name-server-error-1
kmsbehaviourloss: time
medium

When You look up your organisation's external KMS by a name it does not have, through the API or a tool built on it.

Got You get 'Something went wrong', as if the server were broken.

Want You are told no external KMS has that name.

steps
  1. Sign in as an organisation admin and get an access token
  2. Call GET /api/v1/external-kms/name/<a name that matches no external KMS in the organisation>
an external KMS looked up by a name that does not exist is refused as not found
e2e/specs/kms/s11-external-kms-unknown-name-server-error.spec.ts:6
lines
infisical/backend/src/ee/services/external-kms/external-kms-service.ts:433
infisical/backend/src/ee/services/external-kms/external-kms-service.ts:438
before the PRs testedno
Downloading the private key of a certificate that no longer exists crashes with "Something went wrong" instead of saying it was not found
s2-private-key-missing-certificate-crashes-1
behaviourloss: nothing
low

When

Got

Want

steps
  1. Sign in as a user of an organisation
  2. Request the private key of a certificate id that no certificate carries: GET /api/v1/cert-manager/certificates/<random uuid>/private-key
  3. Or by serial number: GET /api/v1/pki/certificates/0/private-key
  4. Observe the answer and the API log
lines
infisical/backend/src/services/certificate/certificate-service.ts:403
infisical/backend/src/services/certificate/certificate-service.ts:408
infisical/backend/src/server/routes/v1/certificate-router.ts:1627
before the PRs testedno
Renaming a project wipes a description another admin just saved
s10-rename-from-open-settings-wipes-new-description-1
projectsbehaviourloss: work
low

When Two admins have the same project's settings open and one saves a new description.

Got When the other renames the project, the new description quietly disappears.

Want The rename changes only the name and the saved description stays.

steps
  1. Create a Secrets Management project
  2. Admin A and admin B (two signed-in tabs of organisation admins) open the project's Settings
  3. Admin A types a new Project description and clicks Save changes
  4. Admin B, without reloading, changes only the Project name and clicks Save changes
  5. Read the project (reload its settings)
renaming a project from a settings page opened earlier keeps the description another admin just saved
e2e/specs/projects/s10-rename-from-open-settings-wipes-new-description.spec.ts:1
lines
infisical/frontend/src/components/project/ProjectOverviewChangeSection.tsx:61
infisical/frontend/src/components/project/ProjectOverviewChangeSection.tsx:72
infisical/frontend/src/hooks/api/projects/queries.tsx:266
infisical/backend/src/services/project/project-service.ts:1055
before the PRs testedno
Returning to OIDC sign-in without a started sign-in crashes the server
s23-oidc-return-without-sign-in-crashes-1
sso-loginbehaviourloss: time
low

When You come back to the OIDC sign-in page without a sign-in in progress, for example by reloading it.

Got You get a bare 'Something went wrong' server error.

Want You are told the OIDC sign-in failed and can start it again.

steps
  1. In a browser that has not started an OIDC sign-in, open the API's OIDC sign-in return address /api/v1/sso/oidc/callback with any code and state
  2. Read the answer
s23-oidc-return-without-sign-in-crashes
e2e/specs/sso-login/s23-oidc-return-without-sign-in-crashes.spec.ts:5
lines
infisical/backend/src/ee/services/email-domain/email-domain-fns.ts:48
infisical/backend/src/ee/routes/v1/oidc-router.ts:128
infisical/backend/src/ee/services/oidc/oidc-config-service.ts:809
before the PRs testedno
Saving one secret from several places at once fails some saves
s8-simultaneous-saves-of-one-secret-fail-1
secrets-managementperformanceloss: work
medium

When Several people or scripts save the same secret at the same moment.

Got Some saves fail with an error after about a minute, and every save in the group waits about a minute.

Want Every save goes through, or is refused clearly, without a long wait.

steps
  1. Create a shared secret in the dev environment of a project through the API and save it once
  2. Send twelve saves of that secret at the same moment, each with a different value
  3. Read the twelve answers
s8-simultaneous-saves-of-one-secret-fail
e2e/specs/secrets-management/s8-simultaneous-saves-of-one-secret-fail.spec.ts:9
lines
infisical/backend/src/services/secret-v2-bridge/secret-v2-bridge-service.ts:807
infisical/backend/src/services/secret/secret-service.ts:1967
infisical/backend/src/server/routes/v4/secret-router.ts:654
infisical/backend/src/db/instance.ts:36
before the PRs testedno
A project's secrets look missing when the overview fails to load
s2-overview-load-failure-shown-as-empty-1
uiloss: time
medium

When

Got

Want

steps
  1. Sign in as a member of an organisation whose secret-management project holds secrets in its environments.
  2. Open the project's Secrets overview while the server fails to answer the secrets overview request.
  3. Look at the secrets table.
lines
infisical/frontend/src/pages/secret-manager/OverviewPage/OverviewPage.tsx:987
infisical/frontend/src/pages/secret-manager/OverviewPage/OverviewPage.tsx:1017
infisical/frontend/src/pages/secret-manager/OverviewPage/OverviewPage.tsx:2850
infisical/frontend/src/pages/secret-manager/OverviewPage/components/TableEmptyRow/TableEmptyRow.tsx:85
infisical/frontend/src/hooks/api/reactQuery.tsx:343
before the PRs testedno
A failed SCIM events load says there are no SCIM events
s1-scim-events-load-error-shown-as-empty-1
behaviourloss: decision
medium

When

Got

Want

steps
  1. Sign in as an admin of an organization with SCIM provisioning enabled and SCIM events recorded in the last 30 days.
  2. Open Organization Settings > SSO & Provisioning > Provisioning: the SCIM Events table lists the events.
  3. Make the SCIM events request fail on the server (a 500 answer).
  4. Reload the Provisioning tab.
lines
infisical/frontend/src/pages/organization/SettingsPage/components/OrgProvisioningTab/ScimEvents.tsx:46
infisical/frontend/src/pages/organization/SettingsPage/components/OrgProvisioningTab/ScimEvents.tsx:33
infisical/frontend/src/hooks/api/scim/queries.tsx:32
before the PRs testedno
SCIM credentials say no token exists when the token list fails to load
s1-scim-token-load-failure-shown-as-empty-1
behaviourloss: decision
medium

When

Got

Want

steps
  1. As an organization admin with SCIM enabled, create a SCIM token in Manage SCIM Credentials.
  2. Make the SCIM token list request fail (the server answers an error).
  3. Open Organization Settings > SSO & Provisioning > Provisioning and click Configure under SCIM to open Manage SCIM Credentials.
lines
infisical/frontend/src/pages/organization/SettingsPage/components/OrgProvisioningTab/ScimTokenModal.tsx:90
infisical/frontend/src/pages/organization/SettingsPage/components/OrgProvisioningTab/ScimTokenModal.tsx:142
infisical/frontend/src/pages/organization/SettingsPage/components/OrgProvisioningTab/ScimTokenModal.tsx:348
infisical/frontend/src/hooks/api/scim/queries.tsx:12
before the PRs testedno
One use of a honey token's key is listed twice in its events
s6-redelivered-honey-token-event-recorded-twice-1
secret-insightsbehaviourloss: decision
medium

When The honey token alert pipeline reports the same use of a planted key twice, which it may do.

Got The token's events show two uses where the key was used once, so the incident looks bigger than it is.

Want Each real use of the key appears once.

steps
  1. Configure the AWS honey token integration for the organisation and plant an AWS honey token in a secrets project.
  2. Have the honey token pipeline report one access to the token's key (one CloudTrail event id), signed with the organisation's webhook signing key.
  3. Open the token's events: the access is listed once.
  4. Have the pipeline report the same access again (same event id, a new signature), as its standard SQS queue may do.
  5. Open the token's events again.
an AWS access event delivered twice appears once in the honey token's events
e2e/specs/secret-insights/s6-redelivered-honey-token-event-recorded-twice.spec.ts:1
lines
infisical/backend/src/ee/services/honey-token/honey-token-service.ts:1126
infisical/backend/src/ee/services/honey-token/honey-token-service.ts:1034
infisical/cloudformation/honey-token-pipeline/honey-tokens-v1.yaml:70
before the PRs testedno
SCIM group sync slows down as the organization's group count grows
s5-scim-group-list-member-query-per-group-1
scim-provisioningperformanceloss: time
medium

When Your identity provider syncs groups into an organization that has many groups.

Got Each sync page gets slower with every group on it, because each group's members are fetched separately.

Want A page of groups takes about the same effort whether it holds 3 groups or 30.

steps
  1. Enable SCIM provisioning for an organization and create a SCIM token
  2. As the identity provider, create 30 SCIM groups with that token
  3. List SCIM groups with a page size of 3, counting the database statements
  4. List SCIM groups with a page size of 30, counting the database statements
listing a page of 30 SCIM groups runs no more member statements than a page of 3
e2e/specs/scim-provisioning/s5-scim-group-list-member-query-per-group.spec.ts:11
lines
infisical/backend/src/ee/services/scim/scim-service.ts:1145
infisical/backend/src/ee/services/scim/scim-service.ts:1146
infisical/backend/src/ee/services/group/user-group-membership-dal.ts:240
before the PRs testedno
Saving SCIM role mappings after a failed load deletes the existing ones
s5-scim-role-mappings-wiped-after-failed-load-1
scim-provisioningbehaviourloss: work
high

When Your organization's SCIM role mappings fail to load when you open the Provisioning tab.

Got The section looks empty with no warning. You add a mapping and save, and every mapping you had before is gone.

Want You are told the mappings could not be loaded, and a save never deletes the ones you had.

steps
  1. As an org admin, save one SCIM group-to-role mapping (Organization Settings > SSO & Provisioning > Provisioning).
  2. Make the request that loads the mappings fail (a server error on the list request).
  3. Open the Provisioning tab: the SCIM Group to Organization Role Mappings section shows no mappings and no error.
  4. Click Add Mapping, type a group name, click Update Mappings.
  5. Reload the mappings.
s5-scim-role-mappings-wiped-after-failed-load
e2e/specs/scim-provisioning/s5-scim-role-mappings-wiped-after-failed-load.spec.ts:12
lines
infisical/frontend/src/pages/organization/SettingsPage/components/OrgProvisioningTab/ExternalGroupOrgRoleMappings.tsx:53
infisical/frontend/src/pages/organization/SettingsPage/components/OrgProvisioningTab/ExternalGroupOrgRoleMappings.tsx:67
infisical/frontend/src/hooks/api/externalGroupOrgRoleMappings/queries.tsx:22
infisical/backend/src/services/external-group-org-role-mapping/external-group-org-role-mapping-service.ts:165
infisical/backend/src/services/external-group-org-role-mapping/external-group-org-role-mapping-dal.ts:49
before the PRs testedno
Reading one SCIM group loads every member of the organization
s5-scim-group-read-loads-every-org-member-1
scim-provisioningperformanceloss: time
low

When Your identity provider syncs a group into an organization with many members.

Got Each time it reads that group, Infisical loads every member of the organization to find the few in the group. The work grows with each member added.

Want Reading a group looks up only that group's members.

steps
  1. Enable SCIM for an organization and create a SCIM token.
  2. Over SCIM, create 2 users and a group with those 2 users as members.
  3. Read the group over SCIM, counting the rows of the organization-membership query.
  4. Provision 40 more users into the organization over SCIM.
  5. Read the same group again, counting the same query.
s5-scim-group-read-loads-every-org-member
e2e/specs/scim-provisioning/s5-scim-group-read-loads-every-org-member.spec.ts:28
lines
infisical/backend/src/ee/services/scim/scim-service.ts:1392
infisical/backend/src/ee/services/scim/scim-service.ts:1403
infisical/backend/src/ee/services/group/group-dal.ts:122
before the PRs testedno
A failed search across folders says no secret matches
s9-secret-search-failure-shown-as-no-matches-1
secrets-managementuiloss: decision
medium

When You search all folders of a project and the search fails.

Got The dialog says "No Matching Resources", so the secret looks like it does not exist.

Want The dialog says the search failed and lets you try again.

steps
  1. Sign in as the shard's user and open the secret-management project's overview
  2. Click the search box and choose "Search all folders"
  3. Make the server fail the folder-wide search request
  4. Type the key of a secret that exists in the project
s9-secret-search-failure-shown-as-no-matches
e2e/specs/secrets-management/s9-secret-search-failure-shown-as-no-matches.spec.ts:9
lines
infisical/frontend/src/pages/secret-manager/OverviewPage/components/SecretSearchInput/components/QuickSearchModal.tsx:275
infisical/frontend/src/pages/secret-manager/OverviewPage/components/SecretSearchInput/components/QuickSearchModal.tsx:296
infisical/frontend/src/pages/secret-manager/OverviewPage/components/SecretSearchInput/components/QuickSearchModal.tsx:546
infisical/frontend/src/pages/secret-manager/OverviewPage/components/SecretSearchInput/components/QuickSearchModal.tsx:253
infisical/frontend/src/pages/secret-manager/OverviewPage/components/SecretSearchInput/components/QuickSearchModal.tsx:513
before the PRs testedno
Saving many secrets at once costs one database write per secret
s9-batch-save-updates-each-secret-separately-1
secrets-managementperformanceloss: time
medium

When You save changes to many secrets at once.

Got The save gets slower with every secret in the batch, since each one is written separately.

Want A batch save takes about the same work whether it holds 5 secrets or 50.

steps
  1. Create 50 shared secrets in one folder of an environment with one batch create.
  2. Save new values for 5 of them in one batch update, three times, counting the database statements each time.
  3. Save new values for all 50 in one batch update, three times, counting the statements each time.
  4. Compare the count of UPDATE statements on the secrets table between the two sizes.
s9-batch-save-updates-each-secret-separately
e2e/specs/secrets-management/s9-batch-save-updates-each-secret-separately.spec.ts:9
lines
infisical/backend/src/services/secret-v2-bridge/secret-v2-bridge-dal.ts:263
infisical/backend/src/services/secret-v2-bridge/secret-v2-bridge-dal.ts:264
infisical/backend/src/services/secret-v2-bridge/secret-v2-bridge-fns.ts:272
infisical/backend/src/server/routes/v4/secret-router.ts:1170
before the PRs testedno
A honey token's security alert email is lost for 24 hours when the API stops right after acknowledging the trigger
s7-honey-token-alert-lost-when-api-crashes-1
secret-insightsbehaviourloss: decision
medium

When Someone uses a honey token's AWS key and the Infisical server stops just after recording it.

Got The token shows as triggered, but no admin ever gets the security alert email, even when the key is used again that day.

Want The admins get the alert email once the server is back.

steps
  1. Plant an AWS honey token in a secrets project of an organisation whose admins receive alerts
  2. Make the mail service slow to accept mail
  3. Have the honey token pipeline report a use of the token's access key (a signed trigger); the API acknowledges it and the token shows as triggered
  4. Kill the API process before the alert email has gone out, then start it again
  5. Use the honey token's access key again (a second signed trigger with a new event id)
  6. Look in the admins' inbox for "Security Alert: Honey Token … Triggered"
s7-honey-token-alert-lost-when-api-crashes: a triggered honey token alerts its org admins even when the API crashed after acknowledging
e2e/specs/secret-insights/s7-honey-token-alert-lost-when-api-crashes.spec.ts:49
lines
infisical/backend/src/ee/services/honey-token/honey-token-service.ts:1131
infisical/backend/src/ee/services/honey-token/honey-token-service.ts:1146
infisical/backend/src/ee/services/honey-token/honey-token-service.ts:980
infisical/backend/src/ee/services/honey-token/honey-token-dal.ts:105
before the PRs testedno
The SCIM events log reads every event of the organization to show one page
s5-scim-events-page-reads-every-event-1
scim-provisioningperformanceloss: time
medium

When Your identity provider has synced users over SCIM for a while and you open the SCIM events log.

Got Each page makes the database read every SCIM event your organization has in the time window, so the log gets slower as events pile up.

Want Opening a page of events costs about a page of work, however many events exist.

steps
  1. Enable SCIM for an organization and create a SCIM token.
  2. Let the identity provider list users over SCIM for a month (about 20,000 SCIM events in the last 30 days; the test copies the product's own list-users event row up to that volume and runs ANALYZE).
  3. As the org admin, load the first page of SCIM events: GET /api/v1/scim/scim-events?since=30d&limit=10&offset=0.
  4. Read the plan of the statement the route runs for that org and window.
s5-scim-events-page-reads-every-event
e2e/specs/scim-provisioning/s5-scim-events-page-reads-every-event.spec.ts:32
lines
infisical/backend/src/ee/services/scim/scim-events-dal.ts:13
infisical/backend/src/ee/services/scim/scim-events-dal.ts:17
infisical/backend/src/db/migrations/20260105152608_scim-events.ts:66
before the PRs testedno
An invited member provisioned twice at once gets two SSO links
s8-scim-provision-twice-duplicate-sso-link-1
scim-provisioningbehaviourloss: time
low

When Your identity provider sends the same new-user request twice at the same moment for someone already invited.

Got Both go through and the person ends up linked to the organization's single sign-on twice.

Want The person has one single sign-on link, whatever the timing.

steps
  1. As an organization admin, configure OIDC single sign-on, verify an email domain, turn on SCIM provisioning and create a SCIM token.
  2. Invite an email on that domain to the organization (the person becomes an invited member).
  3. As the identity provider, send the SCIM create-user request for that email twice at the same moment, with the same user name.
  4. Count the member's SSO links in the organization.
s8-scim-provision-twice-duplicate-sso-link
e2e/specs/scim-provisioning/s8-scim-provision-twice-duplicate-sso-link.spec.ts:39
lines
infisical/backend/src/ee/services/scim/scim-service.ts:468
infisical/backend/src/ee/services/scim/scim-service.ts:538
infisical/backend/src/db/migrations/20260826120000_add-user-alias-lookup-indexes.ts:9
before the PRs testedno
A server failure during MFA sign-in is blamed on the user's code
s10-mfa-server-error-shown-as-invalid-code-1
sign-inbehaviourloss: decision
medium

When You enter the right MFA code from your email, but the server has a problem checking it.

Got The screen says your code is invalid and takes away one of your attempts; enough of these and you are signed out as locked.

Want It says the code could not be checked and lets you try again without losing an attempt.

steps
  1. Sign up a fresh account and turn on email MFA enforcement for its organization
  2. Sign out, then sign in with email and password until the MFA code prompt shows
  3. Read the six-digit code from the email in Mailpit
  4. Make the MFA code check answer with a server error, or drop the connection to it
  5. Enter the correct code
lines
infisical/frontend/src/components/auth/Mfa.tsx:169
infisical/frontend/src/components/auth/Mfa.tsx:172
infisical/frontend/src/components/auth/Mfa.tsx:181
infisical/frontend/src/components/auth/Mfa.tsx:450
infisical/frontend/src/hooks/api/auth/queries.tsx:175
before the PRs testedno
A server failure during sign-in tells the user their password is wrong
s1-server-failure-blames-credentials-1
behaviourloss: decision
medium

When

Got

Want

steps
  1. Open the login page while signed out
  2. Have the login service answer with a server error, drop the sign-in request (no connection), or let it go unanswered until the page gives up (timeout)
  3. Enter a registered account's correct email and password
  4. Press Continue with Email
lines
infisical/frontend/src/pages/auth/LoginPage/components/InitialStep/InitialStep.tsx:267
infisical/frontend/src/pages/auth/LoginPage/components/InitialStep/InitialStep.tsx:310
infisical/frontend/src/components/utilities/attemptLogin.ts:33
before the PRs testedno
A failed password setup sends the user away from the setup page
s3-password-setup-failure-leaves-setup-page-1
behaviourloss: time
medium

When

Got

Want

steps
  1. Sign in to an account and open Personal Settings, Authentication tab.
  2. Press Email setup link and open the password-setup link from the email.
  3. Enter a new password and its confirmation.
  4. Make the set-password request fail on the server (here: a 500 injected on it).
  5. Press Submit.
lines
infisical/frontend/src/pages/auth/PasswordSetupPage/PasswordSetupPage.tsx:98
infisical/frontend/src/pages/auth/PasswordSetupPage/PasswordSetupPage.tsx:99
before the PRs testedno
Changing your password offline doesn't say the connection is down
s3-offline-password-change-not-named-1
behaviourloss: time
low

When

Got

Want

steps
  1. Create an account with an email and password and sign in.
  2. Open Personal Settings, Authentication tab.
  3. Lose the network connection (here: the password-change request aborted as internet disconnected).
  4. Enter the current password and a new password and press Change password.
lines
infisical/frontend/src/pages/user/PersonalSettingsPage/components/ChangePasswordSection/ChangePasswordSection.tsx:100
infisical/frontend/src/pages/user/PersonalSettingsPage/components/ChangePasswordSection/ChangePasswordSection.tsx:102
before the PRs testedno
Signing in from a link to an organization page lands on Projects instead
s4-deep-link-lost-after-sign-in-1
behaviourloss: time
low

When

Got

Want

steps
  1. With no session in the browser, open /organizations/<your org id>/audit-logs (or /access-management).
  2. You are sent to the login page with an 'Access Restricted' notice.
  3. Sign in with your email and password (an account in one organization).
lines
infisical/frontend/src/pages/middlewares/authenticate.tsx:32
infisical/frontend/src/pages/middlewares/authenticate.tsx:36
infisical/frontend/src/helpers/sessionStorage.ts:20
infisical/frontend/src/pages/auth/SelectOrgPage/route.tsx:208
before the PRs testedno
The page-loading bar gives screen readers a role that does not exist
s7-login-loading-bar-invalid-aria-role-1
uiloss: nothing
low

When

Got

Want

steps
  1. Sign in as a user with an organization
  2. Open the login page while signed in, so the app sends you on to your organization
  3. Inspect the thin loading bar that appears at the top of the page during the change of page
lines
infisical/frontend/src/main.tsx:44
infisical/frontend/src/main.tsx:88
before the PRs testedno
A lost connection while entering your organization after OIDC sign-in goes unreported
s24-oidc-org-selection-offline-unreported-1
sso-loginbehaviourloss: time
medium

When You sign in through OIDC and your connection drops just as Infisical opens your only organization.

Got You land on the organization picker with no message, as if you had never been signed in to it.

Want Infisical says that signing in to your organization failed and that the server could not be reached.

steps
  1. Be a member of exactly one organization (seed-org-a) with OIDC single sign-on set up against Keycloak.
  2. Sign out, open the OIDC login page and enter the work email john@oidc.com.
  3. Cut the connection for the organization-selection request (Playwright failRoute 'offline' on **/api/v3/auth/select-organization).
  4. Sign in at Keycloak as john and come back to Infisical.
s24 a lost connection during OIDC organization selection is reported
e2e/specs/sso-login/s24-oidc-org-selection-offline-unreported.spec.ts:10
lines
infisical/frontend/src/pages/auth/SelectOrgPage/route.tsx:224
infisical/frontend/src/pages/auth/SelectOrgPage/route.tsx:238
infisical/frontend/src/pages/auth/SelectOrgPage/SelectOrgPage.tsx:320
before the PRs testedno
The support email on the page-not-found screen is too faint to read easily
s25-not-found-support-link-low-contrast-1
sso-loginuiloss: nothing
low

When You land on the page-not-found screen and look for help.

Got The support email address is a faint grey on a dark background and is hard to read.

Want The support email is readable for people with low vision.

steps
  1. Sign in to Infisical
  2. Open an address the app does not serve, for example /realms/infisical/protocol/openid-connect/auth on the web app
  3. Look at the 'Still stuck? Email support@infisical.com' line under the Back to Home button
  4. Run an accessibility check (axe) on the screen
s25-not-found-support-link-low-contrast
e2e/specs/sso-login/s25-not-found-support-link-low-contrast.spec.ts:9
lines
infisical/frontend/src/pages/public/ErrorPage/components/ErrorPageFrame.tsx:115
infisical/frontend/src/pages/public/ErrorPage/components/ErrorPageFrame.tsx:118
infisical/frontend/src/index.css:137
before the PRs testedno
A project's tags look missing when they fail to load
s2-tag-list-failure-shown-as-empty-1
behaviourloss: decision
medium

When

Got

Want

steps
  1. Create a secret tag in a Secrets Management project.
  2. Make the project's tag list fail to load (the server answers an error).
  3. Open the project's Settings and the Secret Tags tab.
  4. Look at the Secret Tags table.
lines
infisical/frontend/src/hooks/api/tags/queries.tsx:17
infisical/frontend/src/pages/secret-manager/SettingsPage/components/SecretTagsSection/SecretTagsTable.tsx:64
infisical/frontend/src/pages/secret-manager/SettingsPage/components/SecretTagsSection/SecretTagsTable.tsx:123
before the PRs testedno
SCIM user and group lists return the whole organization in one page
s9-scim-list-page-size-uncapped-1
scim-provisioningperformanceloss: time
medium

When Your identity provider asks Infisical for a very large page of users or groups.

Got Infisical sends every user or group in the organization at once, and that answer grows with the organization.

Want Infisical sends at most a fixed number per page and lets the provider page through the rest.

steps
  1. Enable SCIM for an organization and create a SCIM token.
  2. Provision more than 100 users and more than 100 groups over SCIM.
  3. List users over SCIM with startIndex=1 and count=100000.
  4. List groups over SCIM with startIndex=1 and count=100000.
s9-scim-list-page-size-uncapped › ${path}
e2e/specs/scim-provisioning/s9-scim-list-page-size-uncapped.spec.ts:46
lines
infisical/backend/src/ee/routes/v1/scim-router.ts:278
infisical/backend/src/ee/routes/v1/scim-router.ts:539
infisical/backend/src/services/org/org-dal.ts:1116
infisical/backend/src/ee/services/group/group-dal.ts:44
before the PRs testedno
An import looks empty on the project overview when its secrets fail to load
s11-overview-imported-secrets-failure-shown-as-empty-1
folders-imports-tagsuiloss: time
medium

When You open the project overview and the imported secrets fail to load.

Got Expanding an import says "No imported secrets found", as if the source environment had nothing in it.

Want The overview says the imported secrets could not be loaded.

steps
  1. Sign in as a member of an organisation whose secret-management project has a dev folder importing the staging root, and staging holds secrets.
  2. Open the project's Secrets overview at that dev folder while the server fails to answer the imported-secrets request.
  3. Click the staging import row to expand it.
s11-overview-imported-secrets-failure-shown-as-empty
e2e/specs/folders-imports-tags/s11-overview-imported-secrets-failure-shown-as-empty.spec.ts:12
lines
infisical/frontend/src/hooks/api/secretImports/queries.tsx:185
infisical/frontend/src/hooks/api/secretImports/queries.tsx:187
infisical/frontend/src/pages/secret-manager/OverviewPage/OverviewPage.tsx:934
infisical/frontend/src/pages/secret-manager/OverviewPage/components/SecretImportTableRow/SecretImportTableRow.tsx:641
before the PRs testedno
A SCIM deactivation can be undone by a name change arriving at the same time
s9-scim-concurrent-user-updates-undo-deactivation-1
scim-provisioningbehaviourloss: decision
medium

When Your identity provider deactivates a user and updates their name at nearly the same moment.

Got Both updates report success, but the user is active again in Infisical, or the name change is gone. Nothing says so.

Want The user is renamed and stays deactivated.

steps
  1. Enable SCIM for an organization with SSO configured and create a SCIM token
  2. Provision a user over SCIM (POST /api/v1/scim/Users) with givenName Old, active true
  3. Send two PATCH /api/v1/scim/Users/:id requests at the same moment: one replacing active with false, one replacing name.givenName with Renamed (the test forces the overlap by holding the user's membership row lock until both requests are parked on their update, then releasing it)
  4. Both answer success; read the user's first name and the membership's active flag
s9-scim-concurrent-user-updates-undo-deactivation
e2e/specs/scim-provisioning/s9-scim-concurrent-user-updates-undo-deactivation.spec.ts:40
lines
infisical/backend/src/ee/services/scim/scim-service.ts:738
infisical/backend/src/ee/services/scim/scim-service.ts:775
infisical/backend/src/ee/services/scim/scim-service.ts:792
infisical/backend/src/ee/services/scim/scim-service.ts:813
infisical/backend/src/ee/services/scim/scim-service.ts:831
before the PRs testedno
Two folders can end up with the same name in one path
s11-concurrent-folder-rename-duplicates-name-1
folders-imports-tagsbehaviourloss: time
low

When Two people rename two folders in the same place to the same name at the same moment.

Got Both renames succeed and the place now shows two folders with that name.

Want One rename succeeds and the other is told the name is already taken.

steps
  1. In one project environment path, create two folders A and B
  2. Rename A to a name X, then rename B to X: the second rename is refused with "Folder with specified name already exists"
  3. Create two fresh folders C and D in the same path
  4. Rename C and D to the same fresh name Y at the same moment (the test holds writes to the folders table until both requests have passed their name check)
  5. List the folders in that path
s11-concurrent-folder-rename-duplicates-name
e2e/specs/folders-imports-tags/s11-concurrent-folder-rename-duplicates-name.spec.ts:11
lines
infisical/backend/src/services/secret-folder/secret-folder-service.ts:533
infisical/backend/src/services/secret-folder/secret-folder-service.ts:547
infisical/backend/src/services/secret-folder/secret-folder-service.ts:560
infisical/backend/src/services/secret-folder/secret-folder-service.ts:188
infisical/backend/src/db/migrations/20231218092441_secret-folder.ts:7
before the PRs testedno
Two tags in one project can end up with the same slug
s11-concurrent-tag-creates-share-slug-1
folders-imports-tagsbehaviourloss: time
low

When Two people create a tag with the same name in one project at the same moment.

Got Both are saved and the project now has two tags with the same name.

Want One tag is created and the other is refused because the name is taken.

steps
  1. Open a Secrets Management project and create a tag with a new slug, then create a tag with the same slug again: the second is refused, 'Tag already exists'.
  2. Send two creates of another new slug at the same moment (the test holds both inserts until each request has checked the slug).
  3. List the project's tags.
a tag slug stays unique when two creates of it arrive together
e2e/specs/folders-imports-tags/s11-concurrent-tag-creates-share-slug.spec.ts:12
lines
infisical/backend/src/services/secret-tag/secret-tag-service.ts:43
infisical/backend/src/services/secret-tag/secret-tag-service.ts:46
infisical/backend/src/services/secret-tag/secret-tag-service.ts:61
infisical/backend/src/services/secret-tag/secret-tag-service.ts:75
infisical/backend/src/db/migrations/20231218092517_secret-tag.ts:8
before the PRs testedno
Two folders can end up importing each other's secrets
s12-concurrent-opposite-imports-make-cycle-1
folders-imports-tagsbehaviourloss: time
low

When Two people add opposite secret imports between two folders at the same moment.

Got Both imports are saved, so each folder imports the other — a setup the product otherwise refuses as a cyclic import.

Want One import is saved and the other is refused as cyclic.

steps
  1. In one project, create folder P in the dev environment and folder Q in the staging environment.
  2. Add a secret import into dev:/P from staging:/Q and, at the same moment, a secret import into staging:/Q from dev:/P (the test holds the inserts into secret_imports until both requests have run their cyclic-import check).
  3. List the imports of dev:/P and of staging:/Q.
s12-concurrent-opposite-imports-make-cycle
e2e/specs/folders-imports-tags/s12-concurrent-opposite-imports-make-cycle.spec.ts:22
lines
infisical/backend/src/services/secret-import/secret-import-service.ts:229
infisical/backend/src/services/secret-import/secret-import-service.ts:234
infisical/backend/src/services/secret-import/secret-import-service.ts:237
infisical/backend/src/services/secret-import/secret-import-service.ts:364
before the PRs testedno
A secret saved while its folder is being deleted is silently lost
s14-folder-delete-cascades-new-secret-1
folders-imports-tagsbehaviourloss: work
medium

When You delete an empty folder at the same moment a teammate saves a secret into it.

Got Both actions say they worked. The folder is gone, and the teammate's new secret has been deleted with it.

Want The secret survives, or one of you is told the action could not be done.

steps
  1. In a project's dev environment, create a folder and put one secret in it; delete the folder without forcing it — the delete is refused because the folder contains resources.
  2. Create a second, empty folder in the same environment.
  3. Start deleting the empty folder, and hold its delete just after its emptiness check (a share lock on secret_folders parks the DELETE statement).
  4. While the delete waits, save a new secret into that folder — the save answers 200 with the secret.
  5. Let the delete go — it answers 200.
  6. Count the secrets in the deleted folder's id.
s14-folder-delete-cascades-new-secret
e2e/specs/folders-imports-tags/s14-folder-delete-cascades-new-secret.spec.ts:24
lines
infisical/backend/src/services/secret-folder/secret-folder-service.ts:814
infisical/backend/src/services/secret-folder/secret-folder-service.ts:824
infisical/backend/src/db/migrations/20240730181850_secret-v2.ts:24
before the PRs testedno
Adding a folder across environments reports success when it failed in some
s13-overview-folder-partial-failure-shown-as-success-1
folders-imports-tagsbehaviourloss: decision
medium

When You add a folder from the overview and it cannot be created in one of the environments shown.

Got You are told the folder was created and the dialog closes. The folder is missing from that environment and nothing says so.

Want You are told which environments did not get the folder.

steps
  1. Open a secrets project's overview with the dev and staging environments both selected
  2. Choose Add New, then Add Folder, and type a new folder name
  3. Submit while the server fails the folder create for staging only (dev succeeds)
lines
infisical/frontend/src/pages/secret-manager/OverviewPage/OverviewPage.tsx:1468
infisical/frontend/src/pages/secret-manager/OverviewPage/OverviewPage.tsx:1495
infisical/frontend/src/pages/secret-manager/OverviewPage/OverviewPage.tsx:1496
infisical/frontend/src/pages/secret-manager/OverviewPage/OverviewPage.tsx:1500
before the PRs testedno
Opening a folder's secrets reads every project's imports and references
s15-folder-imported-by-lookup-unindexed-1
folders-imports-tagsperformanceloss: time
medium

When You open a folder of secrets, or the project overview.

Got The app looks through every project's imports and references in the whole instance to find what points at this folder. The more projects the instance holds, the slower each folder opens.

Want Finding what imports or references a folder should cost the same however many other projects exist.

steps
  1. Sign in as the shard member and note the shard project's id.
  2. In another project of the same instance, add folders that import from staging, and a staging secret whose value is ${dev.BASE}.
  3. Open the shard project's dev root secrets view (the dashboard asks for its secrets details).
  4. Read the table's index catalog for secret_imports and secret_references_v2, and the plan of the imported-by statement.
a folder's imported-by lookups are served by an index
e2e/specs/folders-imports-tags/s15-folder-imported-by-lookup-unindexed.spec.ts:20
lines
infisical/backend/src/services/secret-import/secret-import-dal.ts:290
infisical/backend/src/services/secret-import/secret-import-dal.ts:302
infisical/backend/src/services/secret-import/secret-import-dal.ts:321
infisical/backend/src/db/migrations/20231218092508_secret-import.ts:12
infisical/backend/src/db/migrations/20240730181850_secret-v2.ts:62
infisical/backend/src/server/routes/v1/dashboard-router.ts:1441
infisical/backend/src/server/routes/v1/dashboard-router.ts:817
before the PRs testedno
The folder list returns every folder in an environment at once, with no paging
s15-folder-list-returns-every-folder-1
folders-imports-tagsperformanceloss: time
medium

When You list the folders of an environment that has many folders, through the API or the CLI.

Got Every folder comes back in one answer, and asking for a smaller page is ignored.

Want Folders come back a page at a time, as the dashboard shows them.

steps
  1. Sign in as a member of a Secret Management project
  2. Create an environment and, in it, a folder 'parent' holding 199 nested folders
  3. Call the folder list for that environment with path '/' and recursive=true
  4. Count the folders in the answer, then call it again with limit=5&offset=0
s15-folder-list-returns-every-folder
e2e/specs/folders-imports-tags/s15-folder-list-returns-every-folder.spec.ts:18
lines
infisical/backend/src/server/routes/v2/secret-folder-router.ts:390
infisical/backend/src/services/secret-folder/secret-folder-service.ts:906
infisical/backend/src/services/secret-folder/secret-folder-service.ts:930
infisical/backend/src/services/secret-folder/secret-folder-dal.ts:386
before the PRs testedno
Moving a folder gets slower with every subfolder it contains
s12-folder-move-commits-per-subfolder-1
folders-imports-tagsperformanceloss: time
medium

When You move a folder that holds many subfolders.

Got The move does extra work for every subfolder, so it gets slower the more subfolders the folder holds.

Want Moving a large folder costs about the same bookkeeping as moving a small one.

steps
  1. Sign in as a member of a project with a dev and a staging environment
  2. In dev, create a folder holding 2 subfolders, and another holding 42 subfolders
  3. Move each folder to the staging root
  4. Count the database statements each move makes
lines
infisical/backend/src/services/secret-folder/secret-folder-service.ts:1993
infisical/backend/src/services/folder-commit/folder-commit-service.ts:921
infisical/backend/src/services/secret-folder/secret-folder-service.ts:1339
before the PRs testedno
A version history that fails to load tells the owner they lack permission
s2-version-history-failure-shown-as-access-denied-1
uiloss: decision
medium

When

Got

Want

steps
  1. Sign in as a project owner and open the secrets dashboard on a folder holding a secret with two versions.
  2. Have the server fail the secret's version list request (a server error or timeout).
  3. Open the secret's row actions and choose Version History.
  4. Wait for the sheet to finish loading.
lines
infisical/frontend/src/pages/secret-manager/OverviewPage/components/SecretTableRow/SecretVersionHistory.tsx:509
infisical/frontend/src/pages/secret-manager/OverviewPage/components/SecretTableRow/SecretVersionHistory.tsx:557
infisical/frontend/src/hooks/api/secrets/queries.tsx:313
before the PRs testedno
Commit History keeps showing the old list after a restore or revert
s9-commit-list-stale-after-restore-1
secret-versioningbehaviourloss: time
medium

When You restore or revert a commit from a folder's Commit History.

Got The history you are taken back to shows the same commits as before, with no new restore commit, until you reload the page.

Want The new commit shows at the top of the history right away.

steps
  1. Open a project's secrets overview on a folder whose secret was created and then updated (2 Commits).
  2. Click [2 Commits] to open Commit History, open the oldest commit.
  3. Open [Restore Options] › [Roll Back to This Commit], type a message, click [Restore] and confirm [Restore].
  4. The sheet returns to Commit History.
a folder's commit history shows the new commit after a restore
e2e/specs/secret-versioning/s9-commit-list-stale-after-restore.spec.ts:9
lines
infisical/frontend/src/hooks/api/folderCommits/queries.tsx:252
infisical/frontend/src/hooks/api/folderCommits/queries.tsx:219
infisical/frontend/src/hooks/api/reactQuery.tsx:351
infisical/frontend/src/components/secrets/CommitHistorySheet/CommitHistorySheet.tsx:80
before the PRs testedno
The login page refuses an email it thinks is invalid without saying why
s8-invalid-email-no-error-on-login-1
uiloss: time
medium

When

Got

Want

steps
  1. Sign out and open the login page.
  2. Type walk.probe@heal-sbx (an email with no domain suffix) into the email field.
  3. Type any password and press Enter, or try to click Continue with Email.
lines
infisical/frontend/src/pages/auth/LoginPage/components/InitialStep/InitialStep.tsx:98
infisical/frontend/src/pages/auth/LoginPage/components/InitialStep/InitialStep.tsx:452
infisical/frontend/src/pages/auth/LoginPage/components/InitialStep/InitialStep.tsx:493
infisical/frontend/src/pages/auth/LoginPage/components/InitialStep/InitialStep.tsx:45
before the PRs testedno
Older commits that fail to load are silently skipped in Commit History
s2-older-commits-failure-not-reported-1
behaviourloss: time
medium

When

Got

Want

steps
  1. Open a folder of a project that holds more than 10 commits (the shard's history folder holds 22).
  2. Open Commit History: the newest 10 commits show with a 'Load 12 older commits' button.
  3. Make the server fail the request for the next page of commits.
  4. Click 'Load 12 older commits'.
lines
infisical/frontend/src/components/secrets/CommitHistorySheet/CommitHistoryView.tsx:138
infisical/frontend/src/components/secrets/CommitHistorySheet/CommitHistoryView.tsx:247
infisical/frontend/src/components/secrets/CommitHistorySheet/CommitHistoryView.tsx:308
infisical/frontend/src/hooks/api/reactQuery.tsx:344
before the PRs testedno
Opening a large commit's details gets slower with every secret it changed
s9-commit-details-read-per-changed-secret-1
secret-versioningperformanceloss: time
medium

When You open the details of a commit that changed many secrets, such as a bulk import.

Got The page does extra work for every secret in the commit, so large commits load slowly.

Want A commit's details load in about the same time whatever number of secrets it changed.

steps
  1. Sign in as the project owner and pick a folder in a secret-manager project
  2. Create 5 secrets in one commit through the batch commit, and 50 secrets in one commit in another folder
  3. Open Commit History and open each commit's details (GET /api/v1/pit/commits/:commitId/changes)
  4. Count the database statements each opening makes
a commit of 50 secrets is read with no more statements per secret than one of 5
e2e/specs/secret-versioning/s9-commit-details-read-per-changed-secret.spec.ts:9
lines
infisical/backend/src/ee/services/pit/pit-service.ts:213
infisical/backend/src/ee/services/pit/pit-service.ts:215
infisical/backend/src/ee/services/pit/pit-service.ts:330
infisical/backend/src/services/secret-v2-bridge/secret-v2-bridge-service.ts:3811
infisical/backend/src/services/secret-v2-bridge/secret-v2-bridge-service.ts:3836
infisical/backend/src/services/folder-commit-changes/folder-commit-changes-dal.ts:82
before the PRs testedno
The sign-up page refuses an email without a domain ending but never says why
s13-signup-invalid-email-unflagged-1
sign-inuiloss: time
medium

When You type an email address missing its ending, like name@company, on the sign-up page.

Got The Continue with Email button just stays grey and pressing Enter does nothing, with no hint what is wrong.

Want The email field is highlighted with a message asking for a valid address.

steps
  1. Open the sign-up page signed out.
  2. Type walk.probe@heal-sbx in the email field.
  3. Press Enter, or try Continue with Email.
s13-signup-invalid-email-unflagged
e2e/specs/sign-in/s13-signup-invalid-email-unflagged.spec.ts:7
lines
infisical/frontend/src/components/auth/InitialSignupStep.tsx:214
infisical/frontend/src/components/auth/InitialSignupStep.tsx:66
infisical/frontend/src/components/auth/InitialSignupStep.tsx:197
before the PRs testedno
After a restore, the dashboard keeps showing the secrets it replaced
s10-dashboard-stale-after-rollback-1
secret-versioninguiloss: time
medium

When You restore a folder to an earlier commit from Commit History and close the sheet.

Got It says the restore completed, but the dashboard still shows the secrets as they were before. Only reloading the page shows the restored folder.

Want The dashboard shows the restored secrets as soon as the restore completes.

steps
  1. In a project folder, create secret ALPHA, then add secret BETA (two commits).
  2. Open the folder on the secrets overview: ALPHA and BETA are listed.
  3. Open Commit History (the "2 Commits" button), open the older commit, Restore Options > Roll Back to This Commit.
  4. Type a message, click Restore, confirm Restore in the dialog.
  5. See "Restore completed successfully", then close the sheet without reloading.
a folder's dashboard drops the secret a roll back removed
e2e/specs/secret-versioning/s10-dashboard-stale-after-rollback.spec.ts:7
lines
infisical/frontend/src/hooks/api/folderCommits/queries.tsx:251
infisical/frontend/src/hooks/api/folderCommits/queries.tsx:218
infisical/frontend/src/components/secrets/CommitHistorySheet/CommitRestoreView.tsx:122
infisical/frontend/src/pages/secret-manager/OverviewPage/OverviewPage.tsx:4308
before the PRs testedno
The commit count on the secrets overview gets slower with every commit in the folder
s10-commit-count-loads-every-commit-1
secret-versioningperformanceloss: time
medium

When You open a project's secrets overview for a folder with a long history.

Got Showing the commit count takes longer the more commits the folder has ever had.

Want The commit count shows just as fast for an old folder as for a new one.

steps
  1. Create a folder in a project and make 5 commits in it, each adding 3 secrets
  2. Create another folder and make 50 commits in it, each adding 3 secrets
  3. Request the commit count of each folder, as the secrets overview does on open
  4. Compare the rows the database returns for each count
a folder's commit count does not load every commit and change
e2e/specs/secret-versioning/s10-commit-count-loads-every-commit.spec.ts:9
lines
infisical/backend/src/services/folder-commit/folder-commit-service.ts:1685
infisical/backend/src/services/folder-commit/folder-commit-dal.ts:35
infisical/backend/src/services/folder-commit/folder-commit-dal.ts:46
infisical/frontend/src/pages/secret-manager/OverviewPage/OverviewPage.tsx:727
before the PRs testedno
Two rollbacks at once give a secret two versions with the same number
s11-concurrent-rollbacks-duplicate-version-1
secret-versioningbehaviourloss: nothing
low

When Two people roll the same folder back to an earlier commit at the same moment.

Got Both succeed and the secret's history shows two versions with the same number.

Want Each version has its own number, or the second roll back is refused.

steps
  1. Create a folder with a secret RB, then update RB once (versions 1 and 2, two commits).
  2. From two sessions, roll the folder back to its first commit at the same time (the test holds RB's row so both rollbacks read its latest version before either writes).
  3. List RB's versions.
s11-concurrent-rollbacks-duplicate-version
e2e/specs/secret-versioning/s11-concurrent-rollbacks-duplicate-version.spec.ts:17
lines
infisical/backend/src/services/folder-commit/folder-commit-service.ts:1063
infisical/backend/src/services/folder-commit/folder-commit-service.ts:1074
infisical/backend/src/services/folder-commit/folder-commit-service.ts:1161
infisical/backend/src/services/folder-commit/folder-commit-service.ts:1198
infisical/backend/src/db/migrations/20240730181850_secret-v2.ts:38
before the PRs testedno
A folder roll back can report success while the secret it restored stays deleted
s11-rollback-racing-delete-leaves-secret-deleted-1
secret-versioningbehaviourloss: decision
medium

When You roll a folder back while a teammate deletes one of its secrets at the same moment.

Got The roll back says it completed and its commit lists the secret as restored, but the secret is gone from the folder.

Want The secret is back in the folder, or the roll back says it could not restore it.

steps
  1. In a folder, create secret S with value v1, then change it to v2.
  2. Open Commit History and pick the first commit (S = v1) to roll back to.
  3. At the same moment, another user deletes S from the dashboard; the delete reaches S first.
  4. The roll back answers success with 1 secret change.
  5. List the folder's secrets.
a roll back racing a delete of its secret restores the secret or is refused
e2e/specs/secret-versioning/s11-rollback-racing-delete-leaves-secret-deleted.spec.ts:9
lines
infisical/backend/src/ee/services/pit/pit-service.ts:428
infisical/backend/src/ee/services/pit/pit-service.ts:496
infisical/backend/src/services/folder-commit/folder-commit-service.ts:1161
infisical/backend/src/services/folder-commit/folder-commit-service.ts:1198
infisical/backend/src/services/folder-commit/folder-commit-service.ts:1498
infisical/backend/src/lib/knex/index.ts:256
before the PRs testedno
Open change requests look gone when the change request list fails to load
s11-change-request-load-failure-shown-as-none-1
secret-approvalsuiloss: time
medium

When You open a project's change requests and the list can't be loaded.

Got It says there are no open change requests, while the tab still counts 23 waiting for review.

Want It tells you the list couldn't be loaded.

steps
  1. Sign in as an approver of a secret-management project whose change policy has open change requests (the shard project holds 23).
  2. Open the project's Approvals page on the Change Requests tab while the server fails to answer the change request list request (the count request still answers).
  3. Look at the Change Requests panel once the page settles.
a failed change-request load is not shown as an empty list
e2e/specs/secret-approvals/s11-change-request-load-failure-shown-as-none.spec.ts:12
lines
infisical/frontend/src/pages/secret-manager/SecretApprovalsPage/components/SecretApprovalRequest/SecretApprovalRequest.tsx:140
infisical/frontend/src/pages/secret-manager/SecretApprovalsPage/components/SecretApprovalRequest/SecretApprovalRequest.tsx:155
infisical/frontend/src/pages/secret-manager/SecretApprovalsPage/components/SecretApprovalRequest/SecretApprovalRequest.tsx:197
infisical/frontend/src/pages/secret-manager/SecretApprovalsPage/components/SecretApprovalRequest/SecretApprovalRequest.tsx:342
before the PRs testedno
Access requests that fail to load look like there are none pending
s2-failed-access-request-load-shown-as-empty-1
behaviourloss: decision
medium

When

Got

Want

steps
  1. Have a pending access request in a project (a member requests temporary access under an access policy).
  2. As an approver, open the project's Approvals page and choose the Access Requests tab.
  3. Let the access-request list fail to load (server error).
  4. Look at the tab.
lines
infisical/frontend/src/pages/secret-manager/SecretApprovalsPage/components/AccessApprovalRequest/AccessApprovalRequest.tsx:230
infisical/frontend/src/pages/secret-manager/SecretApprovalsPage/components/AccessApprovalRequest/AccessApprovalRequest.tsx:253
infisical/frontend/src/pages/secret-manager/SecretApprovalsPage/components/AccessApprovalRequest/AccessApprovalRequest.tsx:696
infisical/frontend/src/hooks/api/accessApproval/queries.tsx:126
infisical/frontend/src/hooks/api/reactQuery.tsx:342
before the PRs testedno
Comparing a commit with now gets slower with every updated secret
s12-commit-compare-reads-versions-per-secret-1
secret-versioningperformanceloss: time
medium

When You open the restore view of a commit before a change that updated many secrets.

Got The comparison does extra work for each updated secret, so it gets slower the more secrets the change touched.

Want The comparison takes about the same work whether 5 or 50 secrets were updated.

steps
  1. In a project, create a folder and add 50 secrets to it in one batch commit.
  2. Update all 50 secrets in a second batch commit.
  3. Open the folder's Commit History, pick the first commit and open its restore view, which compares that commit with the folder's current state.
  4. Count the database statements the comparison runs, and do the same for a folder of 5 secrets.
comparing a commit before 50 updated secrets reads their versions with no more statements per secret than before 5
e2e/specs/secret-versioning/s12-commit-compare-reads-versions-per-secret.spec.ts:12
lines
infisical/backend/src/services/folder-commit/folder-commit-service.ts:660
infisical/backend/src/services/folder-commit/folder-commit-service.ts:686
infisical/backend/src/services/folder-commit/folder-commit-service.ts:664
infisical/backend/src/ee/services/pit/pit-service.ts:322
before the PRs testedno
Rolling a folder back gets slower with every secret it restores
s11-rollback-writes-per-restored-secret-1
secret-versioningperformanceloss: time
medium

When You roll a folder back to an earlier commit that differs in many secrets, such as after a bulk import.

Got The restore does separate work for every secret it puts back, so large restores take longer and longer.

Want A restore takes about the same time whatever number of secrets it puts back.

steps
  1. Sign in as the project owner of a secret-manager project.
  2. In a fresh folder, create 5 secrets in one batch commit, then update all 5 in a second batch commit; do the same with 50 secrets in another folder.
  3. Roll each folder back to its first commit, then to its second, from Commit History (Roll Back to This Commit, deepRollback off).
  4. Count the database statements each roll back makes.
a roll back over 50 secrets writes them with no more statements per secret than over 5
e2e/specs/secret-versioning/s11-rollback-writes-per-restored-secret.spec.ts:11
lines
infisical/backend/src/services/folder-commit/folder-commit-service.ts:1065
infisical/backend/src/services/folder-commit/folder-commit-service.ts:1161
infisical/backend/src/services/folder-commit/folder-commit-service.ts:1184
infisical/backend/src/services/folder-commit/folder-commit-service.ts:1198
infisical/backend/src/services/folder-commit/folder-commit-service.ts:1222
infisical/backend/src/services/folder-commit/folder-commit-service.ts:1226
before the PRs testedno
The restore preview with nested folders gets slower with every subfolder that changed
s11-deep-compare-reads-per-subfolder-1
secret-versioningperformanceloss: time
medium

When You preview restoring a folder to an earlier commit with nested folders included.

Got The preview does separate work for every subfolder that changed since then, so large folder trees take longer to preview.

Want The preview loads in about the same time whatever number of subfolders changed.

steps
  1. Sign in as the project owner and create a folder with 3 subfolders, one secret in each
  2. Note the folder's latest commit, then update the secret in every subfolder
  3. Do the same with another folder holding 30 subfolders
  4. Open each folder's commit in Commit History and open the restore preview including nested folders
  5. Count the database statements each preview makes
a deep compare over 30 changed subfolders makes no more per-folder statements than over 3
e2e/specs/secret-versioning/s11-deep-compare-reads-per-subfolder.spec.ts:11
lines
infisical/backend/src/services/folder-commit/folder-commit-service.ts:1988
infisical/backend/src/services/folder-commit/folder-commit-service.ts:2027
infisical/backend/src/services/folder-commit/folder-commit-service.ts:2122
infisical/backend/src/ee/services/pit/pit-service.ts:311
before the PRs testedno
Version History hides every version older than the newest 100
s11-version-history-hides-versions-past-100-1
secret-versioningbehaviourloss: time
medium

When A secret has been changed more than 100 times.

Got Version History stops at the 100th-newest version with no way to see older ones, so the first versions cannot be viewed or restored there.

Want Version History lets you reach every saved version, down to the first.

steps
  1. In a secrets project, create a secret and update its value 101 times, so it has 102 versions
  2. Open the secrets dashboard on the secret's folder
  3. Open the secret's More secret actions menu and choose Version History
  4. Scroll to the bottom of the version list
a secret's first version stays reachable in Version History past 100 versions
e2e/specs/secret-versioning/s11-version-history-hides-versions-past-100.spec.ts:8
lines
infisical/frontend/src/pages/secret-manager/OverviewPage/components/SecretTableRow/SecretVersionHistory.tsx:514
infisical/backend/src/services/secret-v2-bridge/secret-v2-bridge-service.ts:3123
infisical/backend/src/services/secret-v2-bridge/secret-version-dal.ts:249
before the PRs testedno
Two access policies can end up on the same environment and secret path
s12-concurrent-access-policy-create-duplicates-path-1
secret-approvalsbehaviourloss: time
low

When Two admins add an access policy for the same environment and secret path at the same moment.

Got Both are saved, and the path now has two access policies.

Want One is saved and the other is refused as a duplicate, as it is when they are added one after the other.

steps
  1. Sign in as a project admin of a project with a staging environment.
  2. Create an access policy on staging with a new secret path; create a second one on the same path — it is refused as already existing.
  3. On another new staging path, send two access policy creates at the same moment (forced here by holding writes to the policy table until both requests have passed the duplicate check).
  4. List the project's access policies and filter by that path.
access policies created together on one environment path leave a single policy
e2e/specs/secret-approvals/s12-concurrent-access-policy-create-duplicates-path.spec.ts:13
lines
infisical/backend/src/services/secret-access-approval-global-policy-bridge/secret-access-approval-global-policy-bridge-service.ts:377
infisical/backend/src/services/secret-access-approval-global-policy-bridge/secret-access-approval-global-policy-bridge-service.ts:379
infisical/backend/src/services/secret-access-approval-global-policy-bridge/secret-access-approval-global-policy-bridge-service.ts:303
infisical/backend/src/db/migrations/20261006120000_secret-approval-policy.ts:31
before the PRs testedno
The change requests list returns every request in one page when asked
s12-change-request-list-page-size-uncapped-1
secret-approvalsperformanceloss: time
medium

When An API client asks for a very large page of a project's change requests.

Got Infisical sends every change request in the project at once, and that answer keeps growing as the project collects requests.

Want Infisical sends at most a fixed number per page and lets the client page through the rest.

steps
  1. Create a project with a change policy on prod and the shard user as approver.
  2. Create 110 secrets in prod and edit each one, so 110 change requests are open.
  3. List the project's open change requests with limit=1000 and offset=0.
s12-change-request-list-page-size-uncapped
e2e/specs/secret-approvals/s12-change-request-list-page-size-uncapped.spec.ts:16
lines
infisical/backend/src/ee/routes/v1/secret-approval-request-router.ts:43
infisical/backend/src/ee/services/secret-approval-request/secret-approval-request-service.ts:333
infisical/backend/src/ee/services/secret-approval-request/secret-approval-request-fns.ts:142
infisical/backend/src/ee/services/secret-approval-request/secret-approval-request-dal.ts:902
before the PRs testedno
Two identical access requests sent together both open as pending
s14-concurrent-access-requests-duplicate-pending-1
secret-approvalsbehaviourloss: time
low

When You send the same temporary-access request twice at once (a double click, two tabs, a script retry).

Got Both are accepted and the approver sees two identical pending requests.

Want The second is refused as a duplicate, as it is when you send it again later.

steps
  1. Sign in as a project member under an access policy on an environment (the shard user, approver of its own policy).
  2. Request temporary access to that environment once, then send the identical request again: it is refused as a pending duplicate.
  3. With a fresh set of permissions, send the identical temporary-access request twice so both pass the duplicate check before either is stored (the test holds inserts into the requests table until both have parked).
  4. Open the Access Requests tab or count the pending requests for that note.
two identical access requests sent together open a single pending request
e2e/specs/secret-approvals/s14-concurrent-access-requests-duplicate-pending.spec.ts:44
lines
infisical/backend/src/services/secret-access-approval-global-request-bridge/secret-access-approval-global-request-bridge-service.ts:308
infisical/backend/src/services/secret-access-approval-global-request-bridge/secret-access-approval-global-request-bridge-service.ts:316
infisical/backend/src/services/secret-access-approval-global-request-bridge/secret-access-approval-global-request-bridge-service.ts:323
infisical/backend/src/services/secret-access-approval-global-request-bridge/secret-access-approval-global-request-bridge-dal.ts:36
infisical/backend/src/db/migrations/20251203002657_global-approvals.ts:65
before the PRs testedno
Switching OIDC sign-on off and straight back on leaves it off
s13-oidc-double-toggle-saves-flipped-1
behaviourloss: time
medium

When

Got

Want

steps
  1. Sign in as an organization admin whose organization has OIDC single sign-on configured, and open Organization Settings › SSO & Provisioning.
  2. Note whether Enable OIDC is on or off.
  3. With saving the setting taking a moment to answer (about 800 ms), click the Enable OIDC switch twice in quick succession.
  4. Wait for both saves to answer, then reload the page.
lines
infisical/frontend/src/pages/organization/SettingsPage/components/OrgSsoTab/OrgOIDCSection.tsx:201
infisical/frontend/src/pages/organization/SettingsPage/components/OrgSsoTab/OrgOIDCSection.tsx:84
infisical/frontend/src/hooks/api/oidcConfig/mutations.tsx:64
before the PRs testedno
Two change policies saved at the same moment can both govern one secret path
s12-paired-policy-create-two-policies-one-path-1
secret-approvalsbehaviourloss: decision
low

When Two admins save a change policy for the same environment and secret path at the same moment.

Got Both are saved, so the path ends up with two approval policies, though saving a second one later is refused.

Want One is saved and the other is told a policy for that path already exists.

steps
  1. As a project admin, create a change policy on the Development environment for a fresh secret path, then create a second policy on that same path: the second is refused with "A policy for secret path '…' already exists in environment 'dev'".
  2. On another fresh path, send two change-policy creates for that environment and path at the same moment (the test holds the policy insert in the database until both requests have passed the path check, then lets both go).
  3. Look at the policies listed for that environment and path.
two change policies created together on one environment path leave a single policy
e2e/specs/secret-approvals/s12-paired-policy-create-two-policies-one-path.spec.ts:14
lines
infisical/backend/src/ee/services/secret-change-global-policy-bridge/secret-change-global-policy-bridge-service.ts:174
infisical/backend/src/ee/services/secret-change-global-policy-bridge/secret-change-global-policy-bridge-service.ts:188
infisical/backend/src/ee/services/secret-change-global-policy-bridge/secret-change-global-policy-bridge-fns.ts:144
infisical/backend/src/db/migrations/20261006120000_secret-approval-policy.ts:31
before the PRs testedno
The OIDC login page refuses a mistyped work email without saying why
s15-oidc-invalid-work-email-unexplained-1
uiloss: time
medium

When

Got

Want

steps
  1. Sign out and open the OIDC login page (Log in › Continue with OIDC).
  2. Type john.ldap.com into the Work Email field.
  3. Press Enter, then click outside the field.
lines
infisical/frontend/src/pages/auth/LoginSsoPage/LoginSsoPage.tsx:73
infisical/frontend/src/pages/auth/LoginSsoPage/LoginSsoPage.tsx:107
infisical/frontend/src/pages/auth/LoginSsoPage/LoginSsoPage.tsx:123
before the PRs testedno
Cancelling an OIDC settings edit keeps the unsaved Client ID
s11-oidc-cancel-keeps-discarded-client-id-1
behaviourloss: decision
medium

When

Got

Want

steps
  1. Sign in as an admin of an organization whose OIDC single sign-on is configured with Client ID infisical-dev.
  2. Open Organization Settings > SSO & Provisioning.
  3. Open the OIDC configuration options menu and choose Edit Configuration.
  4. Replace the Client ID with discarded-client-id without saving.
  5. Press Cancel.
  6. Open Edit Configuration again and read the Client ID field.
lines
infisical/frontend/src/pages/organization/SettingsPage/components/OrgSsoTab/OIDCModal.tsx:490
infisical/frontend/src/pages/organization/SettingsPage/components/OrgSsoTab/OIDCModal.tsx:161
before the PRs testedno
The access requests list reads the project's whole request history on every load
s15-access-request-list-reads-whole-history-1
secret-approvalsperformanceloss: time
medium

When You open the Access Requests tab of a project that has collected many access requests over time.

Got Every request the project ever had is loaded each time, even when you filter to one environment.

Want Only the requests you are looking at are loaded, a page at a time.

steps
  1. Create a project with an access approval policy on prod and another on staging, the user as approver.
  2. Request temporary read access on prod once.
  3. Request temporary read access on staging and reject it; repeat until the project holds 5 access requests.
  4. List the project's access requests filtered to prod and note the rows the request-list statement returns.
  5. Repeat the staging request-and-reject until the project holds 50 access requests.
  6. List the access requests filtered to prod again and compare the rows read.
an environment-filtered access-request list does not read every request in the project
e2e/specs/secret-approvals/s15-access-request-list-reads-whole-history.spec.ts:28
lines
infisical/backend/src/services/secret-access-approval-global-request-bridge/secret-access-approval-global-request-bridge-dal.ts:59
infisical/backend/src/services/secret-access-approval-global-request-bridge/secret-access-approval-global-request-bridge-service.ts:107
infisical/backend/src/ee/services/access-approval-request/access-approval-request-service.ts:736
infisical/backend/src/ee/services/access-approval-request/access-approval-request-service.ts:781
infisical/backend/src/ee/routes/v1/access-approval-request-router.ts:146
before the PRs testedno
The LDAP sign-in screen forgets the typed organization and username when you come back
s16-ldap-login-input-lost-after-back-1
behaviourloss: time
low

When

Got

Want

steps
  1. Signed out, open the LDAP login page (/login/ldap).
  2. Type an organization slug (acme-heal-ldap-org) and an LDAP username (john); do not submit.
  3. Click the Back to login arrow and wait for the main login page.
  4. Press the browser's back button to return to the LDAP login page.
lines
infisical/frontend/src/pages/auth/LoginLdapPage/LoginLDAPPage.tsx:30
infisical/frontend/src/pages/auth/LoginLdapPage/LoginLDAPPage.tsx:37
before the PRs testedno
Opening the change request list reads every change request's secrets in the instance
s16-change-request-list-scans-request-secrets-1
secret-approvalsperformanceloss: time
medium

When You open a project's change requests.

Got Every open reads the changed secrets of every change request ever made in the whole instance, so the list gets slower as other projects make changes.

Want Opening the list reads only the change requests it shows.

steps
  1. Sign in as the approver of a project with a change policy on prod.
  2. Edit a secret on prod so a change request opens.
  3. Open the project's Approvals page on the Change Requests tab, three times.
  4. Compare the rows the database read by scanning the change-request secrets table with the rows that table holds, and list that table's indexes.
lines
infisical/backend/src/ee/services/secret-approval-request/secret-approval-request-dal.ts:789
infisical/backend/src/ee/services/secret-approval-request/secret-approval-request-service.ts:337
infisical/backend/src/db/migrations/20240730181850_secret-v2.ts:87
infisical/backend/src/db/migrations/20261002193922_secret-change-requests-validate-and-index.ts:39
before the PRs testedno
Closing the OIDC settings panel makes the saved settings reopen blank
s12-oidc-escape-empties-client-id-1
behaviourloss: decision
medium

When

Got

Want

steps
  1. Sign in as an admin of an organization whose OIDC single sign-on is configured with Client ID infisical-dev.
  2. Open Organization Settings > SSO & Provisioning.
  3. Open the OIDC configuration options menu and choose Edit Configuration; the Client ID shows infisical-dev.
  4. Close the panel with Escape (or its X, or a click outside it) without changing anything.
  5. Open Edit Configuration again and read the Client ID field.
lines
infisical/frontend/src/pages/organization/SettingsPage/components/OrgSsoTab/OIDCModal.tsx:231
infisical/frontend/src/pages/organization/SettingsPage/components/OrgSsoTab/OIDCModal.tsx:129
infisical/frontend/src/pages/organization/SettingsPage/components/OrgSsoTab/OIDCModal.tsx:161
before the PRs testedno
A lease revocation that failed is reported as "Lease revoked"
s5-lease-revoke-failure-shown-as-revoked-1
dynamic-secretsbehaviourloss: decision
medium

When You revoke a dynamic secret lease but its credentials cannot be removed from the database.

Got Infisical says "Lease revoked". The database credentials may still be live.

Want Infisical tells you the revocation failed and offers a retry or force delete.

steps
  1. Create a PostgreSQL dynamic secret and generate a lease from it
  2. In the target database, remove the lease's role (or make the revocation statement fail any other way)
  3. On the secrets dashboard, open View leases on the dynamic secret
  4. Click Revoke lease on that lease, type delete and confirm
s5-lease-revoke-failure-shown-as-revoked
e2e/specs/dynamic-secrets/s5-lease-revoke-failure-shown-as-revoked.spec.ts:10
lines
infisical/frontend/src/pages/secret-manager/SecretDashboardPage/components/DynamicSecretListView/DynamicSecretLease.tsx:98
infisical/frontend/src/pages/secret-manager/SecretDashboardPage/components/DynamicSecretListView/DynamicSecretLease.tsx:107
infisical/backend/src/ee/services/dynamic-secret-lease/dynamic-secret-lease-service.ts:424
infisical/backend/src/ee/services/dynamic-secret/providers/sql-database.ts:369
before the PRs testedno
Two environments in one project can end up with the same slug
s9-concurrent-env-rename-and-create-share-slug-1
projectsbehaviourloss: work
low

When Two people change a project's environments at the same moment, one renaming an environment to a slug while the other creates an environment with it.

Got Both changes are saved and the project now has two environments with the same slug.

Want One change is saved and the other is refused because the slug is taken.

steps
  1. Create a Secrets Management project; it has Development, Staging and Production.
  2. Rename Staging to a new slug, then create an environment with that same slug: the create is refused, 'Environment with slug already exists'.
  3. Within 10 seconds, rename Development to a second new slug and, at the same moment, create an environment 'Raced' with that slug (the test holds this project's row and the Development row until each request has checked the slug).
  4. List the project's environments.
lines
infisical/backend/src/services/project-env/project-env-service.ts:71
infisical/backend/src/services/project-env/project-env-service.ts:76
infisical/backend/src/services/project-env/project-env-service.ts:87
infisical/backend/src/services/project-env/project-env-service.ts:174
infisical/backend/src/services/project-env/project-env-service.ts:190
infisical/backend/src/keystore/keystore.ts:860
infisical/backend/src/db/migrations/20231212110939_project.ts:30
before the PRs testedno
A TOTP lease that fails to generate leaves its dialog spinning forever
s2-totp-lease-failure-stuck-spinner-1
behaviourloss: time
medium

When

Got

Want

steps
  1. Create a TOTP dynamic secret with a static secret key in a project's development environment.
  2. Open the project's secrets dashboard for the development environment.
  3. Make the lease service fail (any server error on creating a lease).
  4. Hover the TOTP dynamic secret's row and choose Generate lease.
  5. Wait for the lease request to fail and look at the lease dialog.
lines
infisical/frontend/src/pages/secret-manager/SecretDashboardPage/components/DynamicSecretListView/CreateDynamicSecretLease.tsx:676
infisical/frontend/src/pages/secret-manager/SecretDashboardPage/components/DynamicSecretListView/CreateDynamicSecretLease.tsx:682
infisical/frontend/src/pages/secret-manager/SecretDashboardPage/components/DynamicSecretListView/CreateDynamicSecretLease.tsx:697
infisical/frontend/src/pages/secret-manager/SecretDashboardPage/components/DynamicSecretListView/CreateDynamicSecretLease.tsx:737
before the PRs testedno
Renaming a secret can silently undo a teammate's edit to a secret that uses it
s10-rename-overwrites-concurrent-edit-of-referencing-secret-1
secrets-managementbehaviourloss: work
medium

When You rename a secret at the moment a teammate saves another secret that uses it.

Got Both saves look successful, but the teammate's secret goes back to its old value.

Want The teammate's new value is kept, with the reference updated to the new name.

steps
  1. In one project, create secret A in dev with any value
  2. Create secret B in staging with value x-${dev.A}
  3. Edit B to y-${dev.A}, then rename A to A2: B reads y-${dev.A2} (the rule holds one after the other)
  4. Lock B's row so the next writes to it wait
  5. Send an edit of B to z-${dev.A2}; once it waits, send a rename of A2 to A3
  6. Release the lock: both answer 200
  7. Read B's raw value
s10-rename-overwrites-concurrent-edit-of-referencing-secret
e2e/specs/secrets-management/s10-rename-overwrites-concurrent-edit-of-referencing-secret.spec.ts:11
lines
infisical/backend/src/services/secret-v2-bridge/secret-v2-bridge-fns.ts:852
infisical/backend/src/services/secret-v2-bridge/secret-v2-bridge-fns.ts:870
infisical/backend/src/services/secret-v2-bridge/secret-v2-bridge-fns.ts:904
infisical/backend/src/services/secret-v2-bridge/secret-v2-bridge-service.ts:847
before the PRs testedno
Moving a secret to two environments at once copies it into both
s10-simultaneous-moves-of-one-secret-copy-it-twice-1
secrets-managementbehaviourloss: time
low

When Two people move the same secret to different environments at the same moment.

Got Both are told the move worked, and the secret now sits in both environments.

Want One move goes through and the other is told the secret is no longer there.

steps
  1. Create a shared secret in the dev environment of a project with no approval policy
  2. Move it to staging, then try to move it to prod: the second move is refused because the secret is no longer in dev (the control)
  3. Create a second secret in dev
  4. Send two moves of it at the same moment, one to staging and one to prod, held so both read dev before either removes the secret
  5. List the secrets of dev, staging and prod
s10-simultaneous-moves-of-one-secret-copy-it-twice
e2e/specs/secrets-management/s10-simultaneous-moves-of-one-secret-copy-it-twice.spec.ts:11
lines
infisical/backend/src/services/secret-v2-bridge/secret-v2-bridge-fns.ts:1650
infisical/backend/src/services/secret-v2-bridge/secret-v2-bridge-fns.ts:1661
infisical/backend/src/services/secret-v2-bridge/secret-v2-bridge-fns.ts:2052
infisical/backend/src/services/secret-v2-bridge/secret-v2-bridge-service.ts:3266
before the PRs testedno
A secret refused for its reminder is saved anyway, and retrying says it exists
s10-create-refused-for-reminder-leaves-secret-1
secrets-managementbehaviourloss: decision
medium

When Someone allowed to create secrets but not edit them adds a new secret with a reminder.

Got They are told they are not allowed, but the secret is saved without its reminder. Trying again says it already exists.

Want Either the secret is created with its reminder, or nothing is saved.

steps
  1. As an org admin, create a secret-management project and a custom project role allowing secrets read and create, not edit (plus secret-folders read)
  2. Create a machine identity with token auth, give it the role on the project, and take its access token
  3. As the identity, create a plain secret in dev — it is created
  4. As the identity, create a new secret in dev with secretReminderRepeatDays 7 — it answers 403 'You are not allowed to edit on secrets'
  5. As the admin, read that secret — it is there
  6. As the identity, send the same create again — it answers 400 'Secret ... already exists'
s10-create-refused-for-reminder-leaves-secret
e2e/specs/secrets-management/s10-create-refused-for-reminder-leaves-secret.spec.ts:10
lines
infisical/backend/src/services/secret-v2-bridge/secret-v2-bridge-service.ts:481
infisical/backend/src/services/secret-v2-bridge/secret-v2-bridge-service.ts:530
infisical/backend/src/services/reminder/reminder-service.ts:178
infisical/backend/src/services/secret-v2-bridge/secret-v2-bridge-service.ts:793
before the PRs testedno
Revealing all secret values on the overview loads each value separately
s11-reveal-all-fetches-each-value-separately-1
secrets-managementperformanceloss: time
medium

When You click Reveal All on a project overview showing one environment.

Got The page loads every secret's value one at a time, so a full page waits on up to a hundred separate loads.

Want All values on the page arrive together, as fast for 50 secrets as for 5.

steps
  1. Create a folder in dev holding 5 secrets and another holding 50.
  2. Open the project overview filtered to the dev environment, inside the 5-secret folder.
  3. Click Reveal All and count the secret value requests the page sends.
  4. Do the same in the 50-secret folder.
s11-reveal-all-fetches-each-value-separately
e2e/specs/secrets-management/s11-reveal-all-fetches-each-value-separately.spec.ts:38
lines
infisical/frontend/src/pages/secret-manager/OverviewPage/OverviewPage.tsx:431
infisical/frontend/src/pages/secret-manager/OverviewPage/OverviewPage.tsx:3481
infisical/frontend/src/pages/secret-manager/OverviewPage/components/SecretTableRow/SecretEditTableRow.tsx:273
infisical/frontend/src/hooks/api/dashboard/queries.tsx:669
infisical/backend/src/server/routes/v1/dashboard-router.ts:1960
before the PRs testedno
Saving a secret's tags can silently remove a tag a teammate just added
s12-overview-tag-save-drops-teammate-tag-1
secrets-managementbehaviourloss: work
low

When You open a secret's tags on the overview while a teammate adds a tag to the same secret.

Got You add your tag and save. The teammate's tag disappears, and nothing tells you.

Want Both tags stay, or you are told the secret changed.

steps
  1. Create tags t1, t2 and t3 in a secret-management project, and a dev secret tagged t1.
  2. Open the project overview, search for the secret, hover its row and open [View secret tags]; the form shows t1.
  3. As a teammate (another session of the project), add tag t2 to the secret.
  4. In the still-open tag form, add t3 and click [Save Tags].
  5. Read the secret's tags.
s12-overview-tag-save-drops-teammate-tag
e2e/specs/secrets-management/s12-overview-tag-save-drops-teammate-tag.spec.ts:10
lines
infisical/frontend/src/pages/secret-manager/OverviewPage/components/SecretTableRow/SecretTagForm.tsx:70
infisical/frontend/src/pages/secret-manager/OverviewPage/components/SecretTableRow/SecretTagForm.tsx:120
infisical/backend/src/services/secret-v2-bridge/secret-v2-bridge-fns.ts:336
infisical/frontend/src/hooks/api/reactQuery.tsx:351
before the PRs testedno
Each project overview page reads every secret in the folder to show 100
s11-overview-page-reads-whole-folder-1
secrets-managementperformanceloss: time
medium

When You open the project overview on a folder that holds many secrets.

Got Each page of 100 secrets makes the database go through every secret in the folder first. The bigger the folder, the slower every page.

Want Showing one page should cost about the same however many secrets the folder holds.

steps
  1. Sign in as the shard member and create a folder in dev of the shard's secret-management project.
  2. Add 1500 secrets to that folder through the batch create API (15 batches of 100).
  3. Load the project overview's first page of 100 secrets for that folder (dev, sorted by name).
  4. Take the paged secrets statement the page ran from pg_stat_statements and EXPLAIN (ANALYZE) it for that folder and the page's rank window.
s11-overview-page-reads-whole-folder
e2e/specs/secrets-management/s11-overview-page-reads-whole-folder.spec.ts:21
lines
infisical/backend/src/services/secret-v2-bridge/secret-v2-bridge-dal.ts:718
infisical/backend/src/services/secret-v2-bridge/secret-v2-bridge-dal.ts:842
infisical/backend/src/services/secret-v2-bridge/secret-v2-bridge-dal.ts:873
infisical/backend/src/services/secret-v2-bridge/secret-v2-bridge-service.ts:1196
infisical/backend/src/server/routes/v1/dashboard-router.ts:722
before the PRs testedno
Deleting many secrets at once costs five database statements per secret
s10-batch-delete-clears-reminders-per-secret-1
secrets-managementperformanceloss: time
medium

When You delete many secrets at once from a project.

Got The delete does extra work for every secret, so a large delete takes longer the more secrets you pick.

Want Deleting a batch does its cleanup once for the whole batch.

steps
  1. Sign in as the shard's org admin and take an org access token
  2. Create 220 shared secrets in dev of the shard's project through the batch create
  3. Delete 5 of them in one batch request, four times, counting the database statements of each of the last three
  4. Delete 50 of them in one batch request, four times, counting the same way
  5. Compare the minimum count per statement shape between the two sizes
lines
infisical/backend/src/services/secret-v2-bridge/secret-v2-bridge-fns.ts:449
infisical/backend/src/services/secret-v2-bridge/secret-v2-bridge-fns.ts:452
infisical/backend/src/services/secret/secret-queue.ts:209
infisical/backend/src/services/reminder/reminder-service.ts:345
infisical/backend/src/services/reminder/reminder-service.ts:346
infisical/backend/src/services/reminder/reminder-service.ts:350
infisical/backend/src/services/secret-v2-bridge/secret-v2-bridge-dal.ts:54
before the PRs testedno
Cancelling a new secret throws away the typed key and value without asking
s13-add-secret-draft-discarded-on-cancel-1
secrets-managementuiloss: time
low

When You start adding a secret, type its name and value, then press Cancel.

Got The form closes straight away and what you typed is gone when you open it again.

Want You are asked whether to discard what you typed, as when editing a secret.

steps
  1. Sign in and open a secret-management project's dev environment secrets screen
  2. Choose Add New, then Add Secret
  3. Type a key and a secret value
  4. Press Cancel
  5. Choose Add New, then Add Secret again
s13-add-secret-draft-discarded-on-cancel
e2e/specs/secrets-management/s13-add-secret-draft-discarded-on-cancel.spec.ts:7
lines
infisical/frontend/src/pages/secret-manager/OverviewPage/OverviewPage.tsx:3973
infisical/frontend/src/pages/secret-manager/OverviewPage/OverviewPage.tsx:3959
infisical/frontend/src/pages/secret-manager/OverviewPage/components/CreateSecretForm/CreateSecretForm.tsx:1035
infisical/frontend/src/pages/secret-manager/OverviewPage/components/SecretTableRow/SecretEditTableRow.tsx:375
before the PRs testedno
Changes sent for review are reported as saved when the policy check fails
s13-batch-commit-sent-for-review-reported-as-saved-1
secrets-managementbehaviourloss: decision
medium

When You save and commit a batch of changes in an environment that needs approval, while the approval policy fails to load.

Got It says your changes were saved and shows the new value, but they only went to review and the old value is still live.

Want It says the changes were sent for review and keeps showing the live value.

steps
  1. Create a secret-management project with a secret in Development and a secret approval policy on Development
  2. Open the project overview filtered to Development with batch mode on, while the approval-policy lookup for Development fails
  3. Edit the secret's value, stage the change, and click Save and Commit
  4. Read the toast, then read the secret's stored value and the project's open approval requests
s13-batch-commit-sent-for-review-reported-as-saved
e2e/specs/secrets-management/s13-batch-commit-sent-for-review-reported-as-saved.spec.ts:11
lines
infisical/frontend/src/pages/secret-manager/OverviewPage/OverviewPage.tsx:734
infisical/frontend/src/pages/secret-manager/OverviewPage/OverviewPage.tsx:740
infisical/frontend/src/pages/secret-manager/OverviewPage/OverviewPage.tsx:2284
infisical/frontend/src/pages/secret-manager/OverviewPage/OverviewPage.tsx:2308
infisical/backend/src/ee/routes/v1/pit-router.ts:704
before the PRs testedno
The secrets page reads every rotation's secret links in the instance to list one folder's rotations
s5-rotation-list-scans-secret-mappings-1
secret-rotationperformanceloss: time
medium

When You open an environment's secrets page in a project that uses secret rotations.

Got Each load reads the secret links of every rotation in the whole instance, so the page gets slower as other projects add rotations.

Want The page reads only the rotations of the folder it shows.

steps
  1. Sign in as a member of a project with a PostgreSQL secret rotation in the Development environment, while other projects in the instance have their own rotations.
  2. Open the Development environment's secrets page three times.
  3. Compare the rows the database read by scanning the rotation-to-secret mapping table with the rows that table holds, and list that table's indexes.
the folder's rotation list reads its mapped secrets through an index on the rotation
e2e/specs/secret-rotation/s5-rotation-list-scans-secret-mappings.spec.ts:11
lines
infisical/backend/src/db/migrations/20250324142105_secret-rotation-v2.ts:48
infisical/backend/src/db/migrations/20260825120200_add-secret-rotation-mapping-secret-id-index.ts:6
infisical/backend/src/ee/services/secret-rotation-v2/secret-rotation-v2-dal.ts:198
infisical/backend/src/ee/services/secret-rotation-v2/secret-rotation-v2-dal.ts:233
before the PRs testedno
Reverting a commit twice at the same moment records two revert commits
s11-concurrent-reverts-record-two-commits-1
secret-versioningbehaviourloss: nothing
low

When Two people revert the same commit in Commit History at the same moment.

Got Both are told the changes were reverted, and the history shows two revert commits for one undo.

Want One revert is recorded; the other is told there is nothing left to revert.

steps
  1. Create a folder holding a secret BASE, then create the secret ALPHA (two commits, the latest creating ALPHA) — or create ALPHA and update it.
  2. Open the latest commit in Commit History, choose Restore Options → Revert Changes and type "revert".
  3. Double-click Revert Changes (or, as two members or in two tabs, confirm the revert at the same moment).
  4. Open the folder's Commit History or read its commit count.
lines
infisical/backend/src/services/folder-commit/folder-commit-service.ts:2294
infisical/backend/src/services/folder-commit/folder-commit-service.ts:2298
infisical/backend/src/services/folder-commit/folder-commit-service.ts:2300
infisical/backend/src/services/folder-commit/folder-commit-service.ts:2308
infisical/backend/src/services/folder-commit/folder-commit-service.ts:2317
infisical/backend/src/services/folder-commit/folder-commit-service.ts:1256
infisical/backend/src/lib/knex/index.ts:300
infisical/frontend/src/components/secrets/CommitHistorySheet/CommitDetailsView.tsx:368
before the PRs testedno
Opening a commit that isn't in the project shows a server error, not "not found"
s13-unknown-commit-answers-server-error-1
secret-versioningbehaviourloss: nothing
low

When You open a commit that does not belong to your project.

Got You get a generic "Something went wrong" error.

Want You are told the commit was not found.

steps
  1. Sign in as a user with a secrets project.
  2. Request the changes of a commit by id in that project, using a well-formed commit id the project does not hold.
  3. Read the answer.
opening a commit the project does not hold answers not found
e2e/specs/secret-versioning/s13-unknown-commit-answers-server-error.spec.ts:7
lines
infisical/backend/src/services/folder-commit/folder-commit-dal.ts:408
infisical/backend/src/services/folder-commit/folder-commit-dal.ts:414
infisical/backend/src/services/folder-commit/folder-commit-service.ts:1707
before the PRs testedno
A project's secret rotation list returns every rotation at once, with no page
s6-rotation-list-has-no-page-size-1
secret-rotationperformanceloss: time
medium

When A project has many secret rotations and a script or the CLI lists them.

Got Every rotation comes back in one answer, so it grows with each rotation added, and asking for a smaller page is ignored.

Want The list comes back a page at a time.

steps
  1. Create a project and 5 PostgreSQL credentials rotations in it
  2. List the project's secret rotations through the API and count the items
  3. Add 100 more rotations to the same project
  4. List the project's secret rotations again, also with limit=10&offset=0, and count the items
listing a project's 105 rotations returns at most a page of 100
e2e/specs/secret-rotation/s6-rotation-list-has-no-page-size.spec.ts:14
lines
infisical/backend/src/ee/routes/v2/secret-rotation-v2-routers/secret-rotation-v2-router.ts:123
infisical/backend/src/ee/services/secret-rotation-v2/secret-rotation-v2-service.ts:391
infisical/backend/src/ee/services/secret-rotation-v2/secret-rotation-v2-dal.ts:164
before the PRs testedno
Leaving project settings throws away an unsaved project name without asking
s7-project-settings-name-lost-on-leave-1
behaviourloss: time
low

When

Got

Want

steps
  1. Open a Secrets Management project's Settings page.
  2. Type a new value in Project name and do not click Save changes.
  3. Click the '<project> Project' link in the header and wait for the project overview to show.
  4. Press the browser's Back button.
lines
infisical/frontend/src/components/project/ProjectOverviewChangeSection.tsx:57
infisical/frontend/src/components/project/ProjectOverviewChangeSection.tsx:59
infisical/frontend/src/hooks/useNavigationBlocker.tsx:41
infisical/frontend/src/pages/secret-manager/OverviewPage/OverviewPage.tsx:1176
#8636
reviewed by greptile, cubic, coderabbit
no
A rotation's generated credentials that fail to load are shown as "No generated credentials found for this rotation"
s1-generated-credentials-load-error-shown-as-none-found-1
behaviourloss: decision
medium

When

Got

Want

steps
  1. Sign in as a member of a project with a PostgreSQL credentials rotation in the Development environment.
  2. Open the Development environment's secrets page.
  3. Make the server fail the rotation's generated-credentials request (a 500).
  4. On the rotation's row, choose View Generated Credentials.
lines
infisical/frontend/src/components/secret-rotations-v2/ViewSecretRotationV2GeneratedCredentials/ViewSecretRotationV2GeneratedCredentials.tsx:66
infisical/frontend/src/components/secret-rotations-v2/ViewSecretRotationV2GeneratedCredentials/ViewSecretRotationV2GeneratedCredentials.tsx:81
infisical/frontend/src/hooks/api/secretRotationsV2/queries.tsx:59
infisical/frontend/src/hooks/api/reactQuery.tsx:344
before the PRs testedno
The secrets page loads every rotation in a folder just to count them
s5-rotation-count-reads-every-rotation-1
secret-rotationperformanceloss: time
medium

When You open an environment's secrets page in a folder that has many secret rotations.

Got To show the rotation total, the page loads every rotation in the folder with all its secrets, even when it shows only one.

Want The page counts the rotations without loading them all.

steps
  1. Create a project and add 3 PostgreSQL credential rotations to its Development root folder
  2. Create a second project and add 30 rotations to its Development root folder
  3. Load each folder's secrets page with rotations included and a page size of one
  4. Count the rows the database returns for the rotation statements during the 30-rotation load
lines
infisical/backend/src/ee/services/secret-rotation-v2/secret-rotation-v2-service.ts:1763
infisical/backend/src/ee/services/secret-rotation-v2/secret-rotation-v2-dal.ts:232
infisical/backend/src/server/routes/v1/dashboard-router.ts:1188
infisical/backend/src/ee/services/secret-rotation-v2/secret-rotation-v2-dal.ts:186
before the PRs testedno
When the list of apps fails to load, the Add Connection panel shows an empty "All apps" section with no error
s2-app-picker-failure-shown-as-empty-1
uiloss: time
medium

When

Got

Want

steps
  1. Sign in as an organisation admin and open Integrations, App Connections tab
  2. Make the request for the list of connectable apps fail with a server error
  3. Click Add Connection
  4. The panel shows the search box and the "All apps" heading with no app cards and no error; searching "AWS" says no matching apps
lines
infisical/frontend/src/pages/organization/AppConnections/AppConnectionsPage/components/AppConnectionList.tsx:64
infisical/frontend/src/pages/organization/AppConnections/AppConnectionsPage/components/AppConnectionList.tsx:68
infisical/frontend/src/pages/organization/AppConnections/AppConnectionsPage/components/AppConnectionList.tsx:113
infisical/frontend/src/hooks/api/appConnections/queries.tsx:36
before the PRs testedno
A failed connection load makes the rotation form look like there are no connections
s7-rotation-connection-picker-load-error-shown-as-none-1
secret-rotationuiloss: time
medium

When You start a new secret rotation and the list of connections can't be loaded.

Got The connection picker only offers to create a connection, as if your project had none, with no error.

Want The picker tells you the connections couldn't be loaded.

steps
  1. Sign in as a user whose project has a PostgreSQL app connection.
  2. Make the server fail the request for the project's available app connections.
  3. Open the project's Development secrets page.
  4. Choose Add New, then Add Secret Rotation, then PostgreSQL Credentials.
  5. Open the PostgreSQL Connection picker.
a rotation's connection picker lists the connections or an error when the available connections fail to load
e2e/specs/secret-rotation/s7-rotation-connection-picker-load-error-shown-as-none.spec.ts:9
lines
infisical/frontend/src/hooks/api/appConnections/queries.tsx:109
infisical/frontend/src/hooks/api/appConnections/queries.tsx:119
infisical/frontend/src/components/secret-rotations-v2/forms/SecretRotationV2ConnectionField.tsx:53
infisical/frontend/src/components/secret-rotations-v2/forms/SecretRotationV2ConnectionField.tsx:171
before the PRs testedno
A refused PostgreSQL rotation still resets both database users' passwords
s7-refused-rotation-changes-db-passwords-1
secret-rotationbehaviourloss: work
high

When You create a PostgreSQL rotation and a secret with the same mapped name is added at that path at the same moment.

Got The rotation fails as 'already exists' and is not saved, but both database logins have already been changed and no one holds the new passwords.

Want Either the rotation is saved, or the two database logins are left exactly as they were.

steps
  1. Create two PostgreSQL users and confirm both sign in with their passwords
  2. Begin creating a PostgreSQL credentials rotation for those two users, mapping a username/password secret whose names are not yet used at the path
  3. While the rotation is changing the two users' passwords on the database, create a secret with the mapped username name at the same path
  4. Let the rotation save finish: it is refused as 'already exists' and no rotation is saved
  5. Try to sign in as the two database users with their original passwords
a postgres rotation refused at save time leaves both DB users' passwords unchanged
e2e/specs/secret-rotation/s7-refused-rotation-changes-db-passwords.spec.ts:1
lines
infisical/backend/src/ee/services/secret-rotation-v2/shared/sql-credentials/sql-credentials-rotation-fns.ts:173
infisical/backend/src/ee/services/secret-rotation-v2/secret-rotation-v2-service.ts:654
infisical/backend/src/ee/services/secret-rotation-v2/secret-rotation-v2-service.ts:670
infisical/backend/src/ee/services/secret-rotation-v2/secret-rotation-v2-service.ts:673
infisical/backend/src/ee/services/secret-rotation-v2/secret-rotation-v2-service.ts:630
before the PRs testedno
App connections look missing when the connection list fails to load
s1-connection-list-failure-shown-as-empty-1
behaviourloss: decision
medium

When

Got

Want

steps
  1. Sign in as an organisation admin whose organisation has app connections (the shard has 20 AWS and 1 PostgreSQL connection).
  2. Make the app connections list request fail with a server error.
  3. Open Integrations, App Connections tab.
lines
infisical/frontend/src/pages/organization/AppConnections/AppConnectionsPage/components/AppConnectionsTable.tsx:102
infisical/frontend/src/pages/organization/AppConnections/AppConnectionsPage/components/AppConnectionsTable.tsx:416
infisical/frontend/src/hooks/api/appConnections/queries.tsx:74
infisical/frontend/src/hooks/api/reactQuery.tsx:343
before the PRs testedno
Provisioning a lease ignores the dynamic secret's default TTL and asks for 1 hour
s6-lease-ignores-default-ttl-1
dynamic-secretsbehaviourloss: time
medium

When You provision a lease from a dynamic secret whose default TTL is shorter than an hour.

Got The form shows 1 hour as the default instead of the secret's own, and the lease is refused as longer than the max TTL.

Want The form starts at the secret's own default TTL and the lease is created.

steps
  1. Create a PostgreSQL dynamic secret in a project's dev environment with default TTL 15m and max TTL 30m
  2. Open the secrets dashboard of that environment
  3. Hover the dynamic secret row and choose Generate lease
  4. Leave the Default TTL field as it is and choose Provision Lease
s6-lease-ignores-default-ttl
e2e/specs/dynamic-secrets/s6-lease-ignores-default-ttl.spec.ts:8
lines
infisical/frontend/src/pages/secret-manager/SecretDashboardPage/components/DynamicSecretListView/CreateDynamicSecretLease.tsx:672
infisical/frontend/src/pages/secret-manager/SecretDashboardPage/components/DynamicSecretListView/CreateDynamicSecretLease.tsx:685
infisical/backend/src/ee/services/dynamic-secret-lease/dynamic-secret-lease-service.ts:145
infisical/frontend/src/pages/secret-manager/SecretDashboardPage/components/DynamicSecretListView/RenewDynamicSecretLease.tsx:58
before the PRs testedno
A schedule edit made during a rotation saves a next rotation date a day too early
s7-schedule-edit-during-rotation-keeps-stale-date-1
secret-rotationbehaviourloss: decision
low

When You change a rotation's interval at the moment its secrets are being rotated.

Got The change saves, but the next rotation is set from before the rotation that just ran, a day earlier than it should be.

Want The next rotation follows the rotation that just ran, or the change asks you to retry.

steps
  1. Create a PostgreSQL credentials rotation with auto-rotation on, interval 1 day, daily rotation time one minute ahead (UTC).
  2. Wait until that minute has passed.
  3. Start an edit that changes the interval to 3 days, and let Rotate Secrets on the same rotation complete after the edit has been checked but before it is saved.
  4. Read the rotation's next rotation date.
s7-schedule-edit-during-rotation-keeps-stale-date
e2e/specs/secret-rotation/s7-schedule-edit-during-rotation-keeps-stale-date.spec.ts:36
lines
infisical/backend/src/ee/services/secret-rotation-v2/secret-rotation-v2-service.ts:791
infisical/backend/src/ee/services/secret-rotation-v2/secret-rotation-v2-service.ts:824
infisical/backend/src/ee/services/secret-rotation-v2/secret-rotation-v2-service.ts:831
infisical/backend/src/ee/services/secret-rotation-v2/secret-rotation-v2-service.ts:882
infisical/backend/src/ee/services/secret-rotation-v2/secret-rotation-v2-service.ts:1419
before the PRs testedno
Dynamic secret lease lists slow down as every organisation's leases grow
s6-lease-list-scans-every-lease-1
dynamic-secretsperformanceloss: time
medium

When You open a dynamic secret's leases or generate a new lease.

Got The product has to look through every lease of every organisation to find yours, so it gets slower as the instance grows.

Want It finds only your secret's leases directly, however many other leases exist.

steps
  1. Sign in as a user with a PostgreSQL dynamic secret in a project's dev environment
  2. Generate a lease on the dynamic secret
  3. Open the dynamic secret's leases
  4. Read the index catalog of the lease table in the database
s6-lease-list-scans-every-lease
e2e/specs/dynamic-secrets/s6-lease-list-scans-every-lease.spec.ts:10
lines
infisical/backend/src/db/migrations/20240318164718_dynamic-secret.ts:43
infisical/backend/src/ee/services/dynamic-secret-lease/dynamic-secret-lease-service.ts:497
infisical/backend/src/ee/services/dynamic-secret-lease/dynamic-secret-lease-service.ts:129
infisical/backend/src/ee/services/dynamic-secret-lease/dynamic-secret-lease-dal.ts:49
before the PRs testedno
Counting a project's access requests loads every request it ever had
s13-pending-access-request-count-reads-every-request-1
secret-approvalsperformanceloss: time
medium

When Your project has built up many access requests over time.

Got Each time the pending count is shown, the product loads every access request the project ever had just to count them, so it gets slower as requests pile up.

Want The count stays as quick with thousands of requests as with five.

steps
  1. Create a project with an access policy on staging, approved by yourself.
  2. File 5 temporary access requests on staging, each with different permissions.
  3. Ask for the project's access request count and note the rows the database returned for it.
  4. File 45 more access requests.
  5. Ask for the count again and compare the rows returned.
lines
infisical/backend/src/services/secret-access-approval-global-request-bridge/secret-access-approval-global-request-bridge-service.ts:162
infisical/backend/src/services/secret-access-approval-global-request-bridge/secret-access-approval-global-request-bridge-service.ts:108
infisical/backend/src/services/approval-policy/approval-request-dal.ts:100
infisical/backend/src/ee/services/access-approval-request/access-approval-request-dal.ts:881
infisical/backend/src/ee/services/access-approval-request/access-approval-request-service.ts:1277
before the PRs testedno
A dynamic secret's leases list loads every lease at once, with no paging
s6-lease-list-unpaged-1
dynamic-secretsperformanceloss: time
medium

When A dynamic secret has built up many leases and you open its leases.

Got Every lease is loaded and drawn at once, so the list grows slower with every lease, up to ten thousand.

Want Leases load a page at a time.

steps
  1. Create a PostgreSQL dynamic secret in a project's dev environment
  2. Generate 60 leases on it
  3. List its leases (the View leases drawer, or the leases list API)
s6-lease-list-unpaged
e2e/specs/dynamic-secrets/s6-lease-list-unpaged.spec.ts:12
lines
infisical/backend/src/ee/routes/v1/dynamic-secret-router.ts:436
infisical/backend/src/ee/services/dynamic-secret-lease/dynamic-secret-lease-service.ts:497
infisical/backend/src/lib/config/env.ts:270
infisical/frontend/src/pages/secret-manager/SecretDashboardPage/components/DynamicSecretListView/DynamicSecretLease.tsx:205
before the PRs testedno
The dynamic secret list slows as the whole instance grows, not just the folder
s7-dynamic-secret-list-unindexed-1
dynamic-secretsperformanceloss: time
medium

When You open a secrets dashboard on an instance that holds many dynamic secrets and metadata entries.

Got The list of the folder's dynamic secrets has to look through every dynamic secret and metadata entry of every organisation, so it gets slower as the instance grows.

Want The list only looks at the folder's own dynamic secrets, however big the instance gets.

steps
  1. Open any project's secrets dashboard for an environment, which lists the folder's dynamic secrets (GET /api/v1/dynamic-secrets?projectSlug&environmentSlug&path=/)
  2. Look at the statement the list runs: dynamic_secrets left join resource_metadata on resource_metadata.dynamicSecretId, where dynamic_secrets.folderId = / in (…)
  3. Look at the index catalog of dynamic_secrets and resource_metadata
s7-dynamic-secret-list-unindexed: dynamic secrets are found by folder through an index
e2e/specs/dynamic-secrets/s7-dynamic-secret-list-unindexed.spec.ts:21
lines
infisical/backend/src/db/migrations/20240318164718_dynamic-secret.ts:26
infisical/backend/src/db/migrations/20250409161555_add-dynamic-secret-to-resource-metadata.ts:8
infisical/backend/src/ee/services/dynamic-secret/dynamic-secret-dal.ts:111
infisical/backend/src/ee/services/dynamic-secret/dynamic-secret-dal.ts:159
infisical/backend/src/ee/services/dynamic-secret/dynamic-secret-dal.ts:74
before the PRs testedno
Two renewals of the same lease both succeed but extend it only once
s7-concurrent-lease-renewals-extend-once-1
dynamic-secretsbehaviourloss: decision
medium

When Two people renew the same dynamic secret lease by an hour at about the same time.

Got Both are told the renewal worked, but the lease is only one hour longer, so its credentials expire an hour earlier than they expect.

Want Each accepted renewal extends the lease, or the second one is refused.

steps
  1. Create a PostgreSQL dynamic secret with default TTL 1h and max TTL 24h, and a lease of it with TTL 1h; note its expiry E
  2. Send two renewals of that lease with TTL 1h at the same time (two members, or an API client), forcing both to read the lease before either writes
  3. Both renewals are answered as successful
  4. Read the lease's expiry
s7-concurrent-lease-renewals-extend-once
e2e/specs/dynamic-secrets/s7-concurrent-lease-renewals-extend-once.spec.ts:12
lines
infisical/backend/src/ee/services/dynamic-secret-lease/dynamic-secret-lease-service.ts:287
infisical/backend/src/ee/services/dynamic-secret-lease/dynamic-secret-lease-service.ts:318
infisical/backend/src/ee/services/dynamic-secret-lease/dynamic-secret-lease-service.ts:327
infisical/backend/src/ee/services/dynamic-secret-lease/dynamic-secret-lease-service.ts:336
before the PRs testedno
Two edits of a dynamic secret's settings at once lose one of the changes
s7-edit-concurrent-settings-change-lost-1
dynamic-secretsbehaviourloss: work
medium

When Two people or scripts change different settings of the same dynamic secret at the same moment.

Got Both are told the change was saved, but one of the changes is gone and the old value is back.

Want Both changes are kept, or one is told to retry.

steps
  1. Create a PostgreSQL dynamic secret through the API.
  2. Send one edit that changes only data.inputs.creationStatement and, at the same time, another that changes only data.inputs.revocationStatement (in the test, both are held at their save by a row lock on the dynamic secret and released together).
  3. Both edits answer 200.
  4. Read the dynamic secret back by name.
s7-edit-concurrent-settings-change-lost: two edits of different settings both survive when their saves meet
e2e/specs/dynamic-secrets/s7-edit-concurrent-settings-change-lost.spec.ts:24
lines
infisical/backend/src/ee/services/dynamic-secret/dynamic-secret-service.ts:302
infisical/backend/src/ee/services/dynamic-secret/dynamic-secret-service.ts:344
infisical/backend/src/ee/services/dynamic-secret/dynamic-secret-service.ts:458
infisical/backend/src/ee/services/dynamic-secret/dynamic-secret-service.ts:466
infisical/backend/src/ee/routes/v1/dynamic-secret-router.ts:166
before the PRs testedno
Renaming a dynamic secret makes the page look it up under its old name
s8-rename-rereads-old-name-1
dynamic-secretsbehaviourloss: nothing
low

When You rename a dynamic secret and save.

Got The rename works and nothing looks wrong, but the page quietly looks the secret up under its old name and is told it does not exist.

Want The page only looks the secret up under its new name.

steps
  1. Open a project's secrets dashboard that holds a dynamic secret
  2. Open Edit dynamic secret on it
  3. Change its name and save
  4. Watch the requests the page makes once the form closes
s8-rename-rereads-old-name
e2e/specs/dynamic-secrets/s8-rename-rereads-old-name.spec.ts:8
lines
infisical/frontend/src/hooks/api/dynamicSecret/mutation.ts:54
infisical/frontend/src/pages/secret-manager/OverviewPage/components/EditDynamicSecretForm/EditDynamicSecretForm.tsx:45
infisical/frontend/src/pages/secret-manager/OverviewPage/OverviewPage.tsx:4077
before the PRs testedno
Listing a project's secret rotations reads its encryption key once per rotation
s5-rotation-list-reads-kms-key-per-rotation-1
secret-rotationperformanceloss: time
medium

When A project has many secret rotations and you open a page that lists them.

Got The list does the project's key lookup again for every rotation, so it gets slower the more rotations the project has.

Want The list costs the same key lookup whether the project has 3 rotations or 23.

steps
  1. Create a project and 3 PostgreSQL credentials rotations in it
  2. List the project's secret rotations five times and count the database statements
  3. Add 20 more rotations to the same project
  4. List the project's secret rotations five times again and count the KMS key lookups
lines
infisical/backend/src/ee/services/secret-rotation-v2/secret-rotation-v2-service.ts:396
infisical/backend/src/ee/services/secret-rotation-v2/secret-rotation-v2-fns.ts:277
infisical/backend/src/services/kms/kms-service.ts:426
infisical/backend/src/services/kms/kms-service.ts:1177
before the PRs testedno
Looking up an organisation's app connections reads every organisation's connections
s5-org-connection-lookup-reads-every-org-1
app-connectionsperformanceloss: time
medium

When You open your organisation's app connections, or a sync or rotation form that offers them.

Got The product looks through every organisation's connections on the instance to find yours, so it slows down as the instance grows.

Want It finds only your organisation's connections directly.

steps
  1. Sign in as an organisation admin whose organisation has app connections, on an instance where other organisations have theirs
  2. Open the org's App Connections tab, or a secret sync or rotation form that lists the available AWS connections
  3. Read the indexes of the app connections table in Postgres
s5-org-connection-lookup-reads-every-org
e2e/specs/app-connections/s5-org-connection-lookup-reads-every-org.spec.ts:12
lines
infisical/backend/src/db/migrations/20250912011133_app-connection-project-id-col.ts:11
infisical/backend/src/db/migrations/20250912011133_app-connection-project-id-col.ts:18
infisical/backend/src/db/migrations/20241218181018_app-connection.ts:17
infisical/backend/src/services/app-connection/app-connection-dal.ts:47
infisical/backend/src/services/app-connection/app-connection-service.ts:1342
before the PRs testedno
Opening App Connections loads every connection of the organisation at once
s5-connection-list-loads-every-connection-1
app-connectionsperformanceloss: time
medium

When Your organisation has many app connections and you open the App Connections tab.

Got The tab loads every connection at once, though it shows only 20 a page, so it gets slower as connections are added.

Want The tab loads only the page you are looking at.

steps
  1. Sign in as an organisation admin whose organisation holds more app connections than one table page (20)
  2. Add 60 AWS app connections
  3. Open Integrations, App Connections tab
  4. Look at the connection list the tab loads, and at the rows the table shows
s5-connection-list-loads-every-connection
e2e/specs/app-connections/s5-connection-list-loads-every-connection.spec.ts:13
lines
infisical/backend/src/server/routes/v1/app-connection-routers/app-connection-router.ts:508
infisical/backend/src/services/app-connection/app-connection-dal.ts:50
infisical/frontend/src/pages/organization/AppConnections/AppConnectionsPage/components/AppConnectionsTable.tsx:102
infisical/frontend/src/pages/organization/AppConnections/AppConnectionsPage/components/AppConnectionsTable.tsx:294
before the PRs testedno
A typed restore message is wiped without warning when going back to Commit Details
s7-restore-message-wiped-by-commit-details-back-1
behaviourloss: time
low

When

Got

Want

steps
  1. Open a project's secrets dashboard on a folder with several commits and open Commit History
  2. Open an older commit (the 4th row) to reach Commit Details
  3. Choose Restore Options → Roll Back to This Commit
  4. Type a reason into the Restore message field
  5. Click the "Commit Details" back link — no warning appears
  6. Choose Restore Options → Roll Back to This Commit again: the Restore message field is empty
lines
infisical/frontend/src/components/secrets/CommitHistorySheet/CommitRestoreView.tsx:71
infisical/frontend/src/components/secrets/CommitHistorySheet/CommitRestoreView.tsx:150
infisical/frontend/src/components/secrets/CommitHistorySheet/CommitHistorySheet.tsx:77
before the PRs testedno
The OAuth return page can load forever with no message
s5-oauth-callback-unhandled-app-loader-forever-1
app-connectionsbehaviourloss: time
low

When You come back from a sign-in at a provider to a link the page cannot match to an app.

Got The page keeps loading with no message and never moves on.

Want It tells you the connection could not be finished and takes you back.

steps
  1. Sign in as an organisation admin.
  2. Open /organizations/<orgId>/app-connections/azure/oauth/callback?code=x&state=abc (the shared Azure return path, state without an app suffix).
  3. Wait 15 seconds.
s5-oauth-callback-unhandled-app-loader-forever
e2e/specs/app-connections/s5-oauth-callback-unhandled-app-loader-forever.spec.ts:5
lines
infisical/frontend/src/pages/organization/AppConnections/OauthCallbackPage/OauthCallbackPage.tsx:63
infisical/frontend/src/pages/organization/AppConnections/OauthCallbackPage/OauthCallbackPage.tsx:680
infisical/frontend/src/pages/organization/AppConnections/OauthCallbackPage/OauthCallbackPage.tsx:708
infisical/frontend/src/pages/organization/AppConnections/OauthCallbackPage/OauthCallbackPage.tsx:714
before the PRs testedno
Commit History forgets your search when you open a commit and go back
s6-commit-search-forgotten-after-opening-commit-1
behaviourloss: time
low

When

Got

Want

steps
  1. Open a project's secrets dashboard on a folder with many commits (the shard's seeded history folder, 22 commits).
  2. Open Commit History.
  3. Type "H0" into Search commits and wait for the list to filter.
  4. Click the first commit to open Commit Details.
  5. Click the "Commit History" back button at the top of the sheet.
lines
infisical/frontend/src/components/secrets/CommitHistorySheet/CommitHistoryView.tsx:131
infisical/frontend/src/components/secrets/CommitHistorySheet/CommitHistoryView.tsx:133
infisical/frontend/src/components/secrets/CommitHistorySheet/CommitHistorySheet.tsx:43
before the PRs testedno
Listing an org's app connections looks up the org key again for every connection
s5-connection-list-reads-org-key-per-connection-1
app-connectionsperformanceloss: time
medium

When Your organisation has many app connections and you open the App Connections tab.

Got The list does the organisation's key lookup again for every connection, so it gets slower the more connections you add.

Want The list costs the same key lookups whether the organisation has 20 connections or 66.

steps
  1. Sign in as an organisation admin whose org already has about 20 app connections
  2. Add 45 more AWS app connections at organisation level (access key method)
  3. List the organisation's app connections three times while counting database statements
listing an org with 45 more connections reads the org key a fixed number of times per list
e2e/specs/app-connections/s5-connection-list-reads-org-key-per-connection.spec.ts:13
lines
infisical/backend/src/services/app-connection/app-connection-service.ts:410
infisical/backend/src/services/app-connection/app-connection-fns.ts:897
infisical/backend/src/services/app-connection/app-connection-fns.ts:534
infisical/backend/src/services/kms/kms-service.ts:928
infisical/backend/src/services/kms/kms-service.ts:377
infisical/backend/src/services/kms/kms-service.ts:426
before the PRs testedno
Secret syncs look missing when the project's sync list fails to load
s1-list-load-failure-shows-empty-1
behaviourloss: decision
medium

When

Got

Want

steps
  1. Sign in as a member of a project that has at least one secret sync (the shard project holds the seeded sync).
  2. Make the project's secret syncs list request fail with a server error.
  3. Open the project's Integrations page on the Secret Syncs tab.
lines
infisical/frontend/src/pages/secret-manager/IntegrationsListPage/components/SecretSyncsTab/SecretSyncsTab.tsx:86
infisical/frontend/src/pages/secret-manager/IntegrationsListPage/components/SecretSyncsTab/SecretSyncTable/SecretSyncsTable.tsx:274
infisical/frontend/src/hooks/api/secretSyncs/queries.tsx:95
infisical/frontend/src/pages/secret-manager/OverviewPage/components/SecretSyncStatusBadgeOverview/SecretSyncStatusBadgeOverview.tsx:46
before the PRs testedno
A secret sync's details page says the sync does not exist when loading it fails
s2-details-load-failure-says-not-found-1
uiloss: time
medium

When

Got

Want

steps
  1. Sign in as the shard user of the secret-syncs project, which has the seeded AWS Parameter Store sync
  2. Make the request that loads that sync's details fail with a server error
  3. Open the sync's details page by its link (Integrations > Secret Syncs > the sync)
  4. Look at the page once loading ends
lines
infisical/frontend/src/pages/secret-manager/SecretSyncDetailsByIDPage/SecretSyncDetailsByIDPage.tsx:53
infisical/frontend/src/pages/secret-manager/SecretSyncDetailsByIDPage/SecretSyncDetailsByIDPage.tsx:61
infisical/frontend/src/hooks/api/secretSyncs/queries.tsx:125
before the PRs testedno
The Secret Syncs tab downloads every sync of a project to show 20
s6-sync-list-loads-every-sync-1
secret-syncsperformanceloss: time
medium

When Your project has many secret syncs and you open the Secret Syncs tab.

Got The page downloads every sync in the project, then shows 20. It downloads them all again every 30 seconds.

Want The page loads only the 20 syncs it shows.

steps
  1. Sign in as a member of a project with more than 20 secret syncs (22 in the repro)
  2. Open the project's Integrations page on the Secret Syncs tab; the table shows 20 per page
  3. Watch the sync list the page downloads, and again every 30 seconds while the tab stays open
the Secret Syncs tab loads at most one page of syncs
e2e/specs/secret-syncs/s6-sync-list-loads-every-sync.spec.ts:7
lines
infisical/backend/src/server/routes/v1/secret-sync-routers/secret-sync-router.ts:248
infisical/backend/src/services/secret-sync/secret-sync-service.ts:383
infisical/backend/src/services/secret-sync/secret-sync-dal.ts:195
infisical/frontend/src/pages/secret-manager/IntegrationsListPage/components/SecretSyncsTab/SecretSyncsTab.tsx:89
infisical/frontend/src/pages/secret-manager/IntegrationsListPage/components/SecretSyncsTab/SecretSyncTable/SecretSyncsTable.tsx:275
before the PRs testedno
Each secrets page load and secret change reads every secret sync in the instance
s5-folder-sync-lookup-scans-all-syncs-1
secret-syncsperformanceloss: time
medium

When You open an environment's secrets page, or change a secret, in any project.

Got Finding which secret syncs use the folder reads every sync in the whole instance, so it slows as other projects add syncs.

Want Only the syncs of that folder are read.

steps
  1. Sign in as a member of a project whose Development root folder is the source of an AWS Parameter Store secret sync, while other projects in the instance hold their own syncs.
  2. Open the Development environment's secrets page for the root folder; it lists the sync as using the folder.
  3. Open the page three more times, comparing the rows the database read by scanning the secret syncs table with the rows the table holds, and list that table's indexes.
the secrets page finds its folder's syncs through an index on the folder
e2e/specs/secret-syncs/s5-folder-sync-lookup-scans-all-syncs.spec.ts:11
lines
infisical/backend/src/db/migrations/20250122055102_secret-sync.ts:21
infisical/backend/src/db/migrations/20250204025010_app-connections-and-secret-syncs-unique-constraint.ts:12
infisical/backend/src/services/secret-sync/secret-sync-service.ts:423
infisical/backend/src/services/secret-sync/secret-sync-queue.ts:1063
infisical/backend/src/services/secret-sync/secret-sync-dal.ts:195
before the PRs testedno
When the subfolder conflict check fails in Add Sync, the Source step says nothing and lets the user continue as if there were no conflicts
s3-conflict-check-failure-shows-nothing-1
behaviourloss: time
medium

When

Got

Want

steps
  1. Open the shard project's Integrations page on the Secret Syncs tab
  2. Click Add Sync and choose AWS Parameter Store
  3. Have the server fail the folder-conflict check (recursive-conflicts answers 500)
  4. On the Source step turn on "Include secrets from all subfolders"
  5. Watch the "Checking for naming conflicts across folders..." spinner clear
lines
infisical/frontend/src/hooks/api/secretSyncs/useRecursiveConflictsCheck.ts:35
infisical/frontend/src/components/secret-syncs/forms/SecretSyncSourceFields.tsx:53
infisical/frontend/src/components/secret-syncs/forms/CreateSecretSyncForm.tsx:251
before the PRs testedno
A failed load of a project's native integrations hides the Native Integrations tab with no error
s1-failed-integrations-load-hides-native-integrations-1
behaviourloss: decision
medium

When

Got

Want

steps
  1. Have a Secrets Management project that already has a native AWS Parameter Store integration
  2. Make the project's integrations read fail with a server error
  3. Open the project's Integrations page on the Native Integrations tab
  4. Observe the page lands on App Connections, the Native Integrations tab is gone and no error is shown
lines
infisical/frontend/src/pages/secret-manager/IntegrationsListPage/route.tsx:81
infisical/frontend/src/pages/secret-manager/IntegrationsListPage/route.tsx:65
infisical/frontend/src/pages/secret-manager/IntegrationsListPage/IntegrationsListPage.tsx:55
before the PRs testedno
After a refused create, the AWS Parameter Store "Create Integration" button keeps spinning
s2-refused-integration-create-leaves-button-spinning-1
uiloss: time
medium

When

Got

Want

steps
  1. Open a Secrets Management project that has an authorized AWS Parameter Store connection
  2. Go to Integrations, choose AWS Parameter Store, and open its Set Up page for that connection
  3. Click Create Integration
  4. The product refuses the create (native integrations are closed to new creation)
  5. Watch the Create Integration button
lines
infisical/frontend/src/pages/secret-manager/integrations/AwsParameterStoreConfigurePage/AwsParamterStoreConfigurePage.tsx:130
infisical/frontend/src/pages/secret-manager/integrations/AwsParameterStoreConfigurePage/AwsParamterStoreConfigurePage.tsx:168
infisical/frontend/src/pages/secret-manager/integrations/AwsParameterStoreConfigurePage/AwsParamterStoreConfigurePage.tsx:342
#8614
reviewed by greptile, cubic, coderabbit
no
An integration's detail page says "Error: Unable to fetch integration." while the integration is still loading
s2-integration-detail-reports-loading-as-fetch-error-1
uiloss: nothing
low

When

Got

Want

steps
  1. Open a Secrets Management project that has a native AWS Parameter Store integration
  2. Open the integration's detail page on a slow connection (the integration takes a few seconds to load)
  3. While it loads, the page shows "Error: Unable to fetch integration."; once the load finishes, the integration's details replace it
lines
infisical/frontend/src/pages/secret-manager/IntegrationsDetailsByIDPage/IntegrationsDetailsByIDPage.tsx:42
infisical/frontend/src/pages/secret-manager/IntegrationsDetailsByIDPage/IntegrationsDetailsByIDPage.tsx:86
infisical/frontend/src/pages/secret-manager/IntegrationsDetailsByIDPage/IntegrationsDetailsByIDPage.tsx:147
before the PRs testedno
Checking a new sync's destination reads every sync of that kind in the organisation
s7-duplicate-check-reads-every-sync-1
secret-syncsperformanceloss: time
low

When You add a secret sync in an organisation that already has many syncs of the same kind.

Got Each check for a duplicate destination reads through every such sync in the organisation, so it gets slower as the organisation grows.

Want The check looks only at syncs that could point at the same place.

steps
  1. Sign in to an organisation that holds many AWS Parameter Store syncs across its projects
  2. Open a project's Integrations page, Secret Syncs tab, and click Add Sync
  3. Pick AWS Parameter Store and type a destination path no sync uses
  4. The form checks the destination for duplicates on each change
checking a new destination for duplicates reads only the syncs that could match
e2e/specs/secret-syncs/s7-duplicate-check-reads-every-sync.spec.ts:13
lines
infisical/backend/src/services/secret-sync/secret-sync-dal.ts:247
infisical/backend/src/services/secret-sync/secret-sync-service.ts:528
infisical/backend/src/services/secret-sync/secret-sync-service.ts:553
infisical/frontend/src/hooks/api/secretSyncs/useDuplicateDestinationCheck.ts:36
before the PRs testedno
Turning auto-sync off can silently undo another member's saved sync options
s7-auto-sync-toggle-reverts-sync-options-1
secret-syncsbehaviourloss: work
medium

When You turn off auto-sync on a sync while a teammate is saving that sync's options.

Got Both saves say they worked, but the teammate's options are quietly put back to what they were.

Want Turning off auto-sync changes only auto-sync and keeps the options your teammate saved.

steps
  1. Create an AWS Parameter Store sync with auto-sync on and "Disable secret deletion" off.
  2. Member A saves the sync's options with "Disable secret deletion" on.
  3. At the same moment member B turns the sync's auto-sync off from its details page, so both saves read the sync before either is written (forced in the test by a row lock on the sync, A parked first, then B).
  4. Both saves report success; reopen the sync.
turning auto-sync off while the sync options are saved keeps both changes
e2e/specs/secret-syncs/s7-auto-sync-toggle-reverts-sync-options.spec.ts:26
lines
infisical/backend/src/services/secret-sync/secret-sync-service.ts:772
infisical/backend/src/services/secret-sync/secret-sync-service.ts:824
infisical/backend/src/services/secret-sync/secret-sync-service.ts:912
infisical/backend/src/services/secret-sync/secret-sync-service.ts:969
infisical/backend/src/services/secret-sync/secret-sync-dal.ts:159
infisical/frontend/src/pages/secret-manager/SecretSyncDetailsByIDPage/components/SecretSyncActionTriggers.tsx:88
before the PRs testedno
Deleting a native integration from its detail page lands on a "Not Found" page instead of the integrations list
s2-delete-from-detail-page-lands-on-not-found-1
uiloss: time
low

When

Got

Want

steps
  1. Open a Secrets Management project that has a native integration (e.g. AWS Parameter Store) and go to Integrations → Native Integrations
  2. Open the integration's detail page
  3. Open the More menu and choose Delete Integration
  4. Type the confirmation text and press Delete
  5. Observe: the 'Deleted integration' toast shows, then the browser lands on /secret-manager/<projectId>/integrations — Infisical's '404 · Not Found' page
lines
infisical/frontend/src/pages/secret-manager/IntegrationsDetailsByIDPage/IntegrationsDetailsByIDPage.tsx:67
infisical/frontend/src/routes.ts:21
infisical/frontend/src/routes.ts:518
before the PRs testedno
The block on duplicate sync destinations lets two syncs to one destination through
s8-duplicate-destination-race-accepts-both-1
secret-syncsbehaviourloss: time
low

When Your organisation blocks duplicate sync destinations and two people add a sync to the same destination at the same moment.

Got Both syncs are created, so two syncs now push to the one destination.

Want One sync is created and the other is refused as a duplicate destination.

steps
  1. Turn on 'Block duplicate secret sync destinations' for the organisation.
  2. Create an AWS app connection.
  3. Create an AWS Parameter Store sync to region us-east-1, path /p/ through that connection, then create a second one to the same path: it is refused as a duplicate destination.
  4. On a new path, send two create requests for syncs with the same region and path through that connection at the same moment (forced in the test by holding the connection's row so both inserts wait after their duplicate checks).
  5. Count the syncs whose destination path is the new path.
two syncs to one destination created together leave one sync when duplicates are blocked
e2e/specs/secret-syncs/s8-duplicate-destination-race-accepts-both.spec.ts:14
lines
infisical/backend/src/services/secret-sync/secret-sync-service.ts:528
infisical/backend/src/services/secret-sync/secret-sync-service.ts:696
infisical/backend/src/services/secret-sync/secret-sync-service.ts:708
infisical/backend/src/services/secret-sync/secret-sync-service.ts:742
infisical/backend/src/db/migrations/20250204025010_app-connections-and-secret-syncs-unique-constraint.ts:12
before the PRs testedno
Listing a project's native integration authorizations reads every organisation's
s9-integration-authorizations-lookup-reads-every-org-1
legacy-integrationsperformanceloss: time
medium

When You open a project's Integrations page, or connect or remove a native integration.

Got The product looks through every organisation's integration authorizations on the instance to find your project's, so it slows as the instance grows.

Want It finds only your project's authorizations directly.

steps
  1. Sign in as a member of a Secrets Management project that has a native AWS Parameter Store authorization, on an instance where other projects have theirs
  2. Open the project's Integrations page (it lists the project's authorizations)
  3. Read the indexes of the integration authorizations table in Postgres
s9-integration-authorizations-lookup-reads-every-org
e2e/specs/legacy-integrations/s9-integration-authorizations-lookup-reads-every-org.spec.ts:11
lines
infisical/backend/src/db/migrations/20231222172455_integration.ts:28
infisical/backend/src/services/integration-auth/integration-auth-service.ts:111
infisical/backend/src/services/integration-auth/integration-auth-service.ts:259
infisical/backend/src/services/integration-auth/integration-auth-service.ts:1829
infisical/backend/src/server/routes/v1/project-router.ts:904
before the PRs testedno
Leaving the secrets page throws away an unsaved secret value without asking
s17-inline-secret-edit-lost-on-leave-1
secret-approvalsbehaviourloss: time
low

When You type a new value for a secret but leave for the Approvals page before saving.

Got Nothing warns you. When you come back, the old value is there and what you typed is gone.

Want The page asks before leaving, or keeps what you typed.

steps
  1. Open a Secrets Management project's secrets page for the prod environment and find a secret.
  2. Click its value box and type a new value; do not click Save and Commit.
  3. Click Approvals in the project navigation and wait for the Approvals page.
  4. Press the browser's Back button and click the secret's value box.
an unsaved inline secret edit is warned about or kept when leaving for Approvals
e2e/specs/secret-approvals/s17-inline-secret-edit-lost-on-leave.spec.ts:14
lines
infisical/frontend/src/pages/secret-manager/OverviewPage/OverviewPage.tsx:1176
infisical/frontend/src/hooks/useNavigationBlocker.tsx:41
infisical/frontend/src/components/v3/platform/SecretInput/SecretInputActions.tsx:28
#8636
reviewed by greptile, cubic, coderabbit
no
Deleting a dynamic secret with many leases fails and leaves leases behind
s9-delete-with-many-leases-fails-1
dynamic-secretsperformanceloss: time
medium

When You delete a database dynamic secret that has many active leases.

Got The deletion ends as "Failed to delete" and some of the leases' database users are still in your database.

Want The dynamic secret is deleted and all its database users removed, however many leases it has.

steps
  1. In a PostgreSQL target database, create a root role with CREATEROLE and CONNECTION LIMIT 20
  2. Create a SQL dynamic secret in Infisical using that root role
  3. Generate 40 leases from it
  4. While the lease roles are briefly busy in the target (an uncommitted ALTER ROLE on each, as an in-flight renewal does), delete the dynamic secret (not forced)
  5. Once the deletion has ended, read the dynamic secret's status and list the lease roles left in the target database
lines
infisical/backend/src/ee/services/dynamic-secret-lease/dynamic-secret-lease-queue.ts:213
infisical/backend/src/ee/services/dynamic-secret-lease/dynamic-secret-lease-queue.ts:221
infisical/backend/src/ee/services/dynamic-secret/providers/sql-database.ts:351
infisical/backend/src/ee/services/dynamic-secret/providers/sql-database.ts:378
before the PRs testedno
Two updates to a native integration's options at once lose one of them
s9-concurrent-integration-option-updates-lose-one-1
legacy-integrationsbehaviourloss: work
medium

When Two scripts or API clients change different options of the same native integration at the same moment.

Got Both are told the change was saved, but one option is gone.

Want Both options are kept, or one is told to retry.

steps
  1. In a Secrets Management project, take an existing native AWS Parameter Store integration with no secret prefix or suffix set.
  2. Send an update that sets only metadata.secretPrefix to "P_" and, at the same moment, one that sets only metadata.secretSuffix to "_S" (in the test both are held at their save by a row lock on the integration and released together).
  3. Both updates answer 200.
  4. Read the integration's metadata back.
s9-concurrent-integration-option-updates-lose-one: two option updates both survive when their saves meet
e2e/specs/legacy-integrations/s9-concurrent-integration-option-updates-lose-one.spec.ts:22
lines
infisical/backend/src/services/integration/integration-service.ts:85
infisical/backend/src/services/integration/integration-service.ts:115
infisical/backend/src/services/integration/integration-service.ts:125
infisical/backend/src/server/routes/v1/integration-router.ts:118
before the PRs testedno
Listing the organisation's integration authorizations slows with every project
s10-org-authorizations-list-checks-permission-per-project-1
legacy-integrationsperformanceloss: time
medium

When You connect a native integration and Infisical offers your existing authorizations to reuse.

Got The list does extra work for every project in the organisation, even projects that hold no authorization for that provider.

Want The list costs the same however many projects the organisation has.

steps
  1. Sign in as an organisation admin with one project holding an AWS Parameter Store authorization
  2. List the organisation's integration authorizations (the select-authorization step of connecting a native integration) and watch the permission-cache reads it makes
  3. Create ten more projects, each with an AWS Parameter Store authorization
  4. List the organisation's integration authorizations again and watch the permission-cache reads
s10-org-authorizations-list-checks-permission-per-project: the org's authorization list reads the permission cache the same with 1 or 11 projects
e2e/specs/legacy-integrations/s10-org-authorizations-list-checks-permission-per-project.spec.ts:31
lines
infisical/backend/src/services/integration-auth/integration-auth-service.ts:116
infisical/backend/src/services/integration-auth/integration-auth-service.ts:121
infisical/backend/src/services/integration-auth/integration-auth-dal.ts:31
infisical/frontend/src/pages/secret-manager/integrations/SelectIntegrationAuthPage/SelectIntegrationAuthPage.tsx:44
before the PRs testedno
A member's details page says the user is not a member of any projects when their projects fail to load
s3-member-projects-load-failure-shown-as-none-1
behaviourloss: decision
medium

When

Got

Want

steps
  1. Sign in as an organization admin and create a project, so the admin belongs to it
  2. Open Access Control > Users, then open that member's details page; the Projects card lists the project
  3. Make the request that lists the member's project memberships fail with a server error, and reload the page
  4. Look at the Projects card
lines
infisical/frontend/src/pages/organization/UserDetailsByIDPage/components/UserProjectsSection/UserProjectsTable.tsx:76
infisical/frontend/src/pages/organization/UserDetailsByIDPage/components/UserProjectsSection/UserProjectsTable.tsx:157
infisical/frontend/src/hooks/api/users/queries.tsx:230
before the PRs testedno
A member's details page is left blank, with no message and no back link, when the member cannot be loaded
s2-member-detail-load-failure-blank-1
uiloss: time
medium

When

Got

Want

steps
  1. Sign in as an organization admin and invite a member
  2. Remove that member from the organization
  3. Open the removed member's details page (/organizations/<orgId>/members/<membershipId>), e.g. from a saved link or browser history
  4. Observe the content area: empty — no member name, no back link, no error or not-found message
  5. Same result when the member exists but the member read fails with a server error
lines
infisical/frontend/src/pages/organization/UserDetailsByIDPage/UserDetailsByIDPage.tsx:50
infisical/frontend/src/pages/organization/UserDetailsByIDPage/UserDetailsByIDPage.tsx:100
infisical/frontend/src/hooks/api/users/queries.tsx:214
before the PRs testedno
Opening the Users tab loads every member of the organization at once
s9-users-list-loads-every-member-1
org-membersperformanceloss: time
medium

When Your organization has many members and you open the Users tab.

Got The tab loads every member at once though it shows 20 a page, so it gets slower with every member added.

Want The tab loads only the page of members you are looking at.

steps
  1. Sign in as an organization admin whose organization holds more members than one Users table page (20)
  2. Invite more members by email
  3. Open Access Management, Users tab
  4. Look at the member list the tab loads, and at the rows the table shows
s9-users-list-loads-every-member
e2e/specs/org-members/s9-users-list-loads-every-member.spec.ts:8
lines
infisical/backend/src/server/routes/v1/organization-router.ts:415
infisical/backend/src/services/org/org-service.ts:289
infisical/backend/src/services/org/org-dal.ts:512
infisical/frontend/src/hooks/api/users/queries.tsx:161
infisical/frontend/src/pages/organization/AccessManagementPage/components/OrgMembersTab/components/OrgMembersSection/OrgMembersTable.tsx:302
before the PRs testedno
Two role changes at once leave a member with both roles, Admin kept
s10-simultaneous-role-changes-leave-two-roles-1
org-membersbehaviourloss: decision
medium

When Two admins change the same member's role at the same moment, one to Admin and one to Member.

Got Both saves succeed, and the member ends up holding both roles — shown as Admin.

Want The member ends with one role, the one saved last.

steps
  1. Sign in as an org admin and invite a new member with role Member.
  2. Send two role changes for that member at the same moment: one to Admin, one to Member (two admins, or two tabs, saving the member detail page's role).
  3. Both saves report success.
  4. Read the member's org roles.
two simultaneous role changes of one member leave exactly one org role
e2e/specs/org-members/s10-simultaneous-role-changes-leave-two-roles.spec.ts:15
lines
infisical/backend/src/services/org/org-service.ts:1038
infisical/backend/src/services/org/org-service.ts:1039
infisical/backend/src/services/org/org-service.ts:1013
infisical/backend/src/services/org/org-service.ts:1020
infisical/backend/src/db/migrations/20251005152640_simplify-membership.ts:130
before the PRs testedno
A refused organization invite still creates an account for the invitee
s11-refused-invite-leaves-user-account-1
org-membersbehaviourloss: nothing
low

When You invite someone with a role your organization does not have.

Got The invite is refused, yet an account for that email is created anyway and stays, belonging to no organization.

Want A refused invite should leave no account behind.

steps
  1. Sign in as an organization admin
  2. Invite an email address that has no Infisical account, choosing an organization role the organization does not have (for example a custom role deleted meanwhile)
  3. See the invite refused: the role was not found
  4. Look up accounts with that email
an invite refused for an unknown role leaves no user account behind
e2e/specs/org-members/s11-refused-invite-leaves-user-account.spec.ts:12
lines
infisical/backend/src/services/membership-user/membership-user-service.ts:203
infisical/backend/src/services/membership-user/membership-user-service.ts:316
infisical/backend/src/services/membership-user/membership-user-service.ts:333
infisical/backend/src/services/membership-user/membership-user-service.ts:371
infisical/backend/src/services/membership-user/membership-user-service.ts:377
infisical/backend/src/ee/services/permission/permission-service.ts:1065
before the PRs testedno
A failed load of a project's groups tells the admin "No project groups found"
s1-failed-groups-load-shown-as-no-groups-1
uiloss: decision
medium

When

Got

Want

steps
  1. Attach an organisation group to a project with the Viewer role.
  2. Make the request listing the project's groups fail with a server error.
  3. Open the project's Access Management page on the Groups tab.
  4. The tab reads "No project groups found — Add a group to get started." and no error appears.
lines
infisical/frontend/src/pages/project/AccessControlPage/components/GroupsTab/components/GroupsSection/GroupsTable.tsx:90
infisical/frontend/src/pages/project/AccessControlPage/components/GroupsTab/components/GroupsSection/GroupsTable.tsx:135
infisical/frontend/src/hooks/api/projects/queries.tsx:561
before the PRs testedno
Removing several members at the same time fails some removals
s10-simultaneous-removals-of-different-members-fail-1
org-membersperformanceloss: work
medium

When Several members are removed from the organization at the same moment, by several admins or a script.

Got Some removals fail with an error after about a minute, and those people stay in the organization.

Want Every removal goes through promptly.

steps
  1. Sign in as an organization admin and invite thirteen new members by email
  2. Remove one of them on its own; it goes through
  3. Send the removals of the other twelve members at the same moment
  4. Read the twelve answers and the member list
s10-simultaneous-removals-of-different-members-fail
e2e/specs/org-members/s10-simultaneous-removals-of-different-members-fail.spec.ts:12
lines
infisical/backend/src/services/org/org-fns.ts:56
infisical/backend/src/services/org/org-fns.ts:112
infisical/backend/src/services/org/org-fns.ts:174
infisical/backend/src/db/instance.ts:36
before the PRs testedno
A member's page says they belong to no groups when their groups fail to load
s9-member-groups-load-failure-shown-as-none-1
org-membersbehaviourloss: decision
medium

When You open a member's details page and their groups cannot be loaded.

Got The page says the user has not been assigned to any groups, with no error, even though they are in one.

Want The page says the groups could not be loaded and lets you retry.

steps
  1. Sign in as an organization admin, create a group and add yourself to it
  2. Open Access Control > Users, then open your own member details page; the groups list returns the new group
  3. Make the request that lists the member's groups fail with a server error, and reload the page
  4. Look at the Groups section
a failed load of a member's groups does not claim the member is in no groups
e2e/specs/org-members/s9-member-groups-load-failure-shown-as-none.spec.ts:14
lines
infisical/frontend/src/pages/organization/UserDetailsByIDPage/components/UserProjectsSection/UserGroupsTable.tsx:42
infisical/frontend/src/pages/organization/UserDetailsByIDPage/components/UserProjectsSection/UserGroupsTable.tsx:132
infisical/frontend/src/hooks/api/users/queries.tsx:477
before the PRs testedno
An organization's member list looks empty when it fails to load
s7-org-users-load-error-shown-as-empty-1
scim-provisioninguiloss: decision
medium

When The list of your organization's members cannot be loaded.

Got The Users tab says no organization users were found and invites you to add some, with no sign anything went wrong.

Want The page says the members could not be loaded and lets you try again.

steps
  1. Sign in as an org admin of an organization that has at least one member (here one provisioned over SCIM).
  2. Make the request that loads the organization's users fail on the server (the first try and its automatic retry).
  3. Open Organization Access Control, Users tab.
lines
infisical/frontend/src/pages/organization/AccessManagementPage/components/OrgMembersTab/components/OrgMembersSection/OrgMembersTable.tsx:124
infisical/frontend/src/pages/organization/AccessManagementPage/components/OrgMembersTab/components/OrgMembersSection/OrgMembersTable.tsx:359
infisical/frontend/src/hooks/api/users/queries.tsx:173
before the PRs testedno
A failed load of a project's machine identities shows a blank table with no error
s2-failed-identities-load-shown-as-blank-table-1
uiloss: decision
medium

When

Got

Want

steps
  1. Attach an organisation machine identity to a project with the Viewer role.
  2. Make the request listing the project's machine identities fail with a server error.
  3. Open the project's Access Management page on the Machine Identities tab.
  4. The table shows only its Name / Project Role / Managed by header: no rows, no empty-state text, no pager and no error.
lines
infisical/frontend/src/pages/project/AccessControlPage/components/IdentityTab/IdentityTab.tsx:144
infisical/frontend/src/pages/project/AccessControlPage/components/IdentityTab/IdentityTab.tsx:291
infisical/frontend/src/pages/project/AccessControlPage/components/IdentityTab/IdentityTab.tsx:626
before the PRs testedno
Moving a rotation can fail with "Something went wrong" instead of a name conflict
s8-rotation-move-race-answers-server-error-1
secret-rotationbehaviourloss: time
low

When You move a rotation to a folder while a teammate creates a rotation there with the same secret names.

Got The move fails with "Something went wrong" and no reason.

Want The move is refused with a message naming the secrets that already exist there.

steps
  1. In one project, create folders /a and /d in Development.
  2. Create PostgreSQL rotation R1 in /a mapping secrets RACE_U and RACE_P.
  3. Start moving R1 to /d.
  4. While the move has checked /d but not yet moved the secrets, create rotation R2 in /d mapping RACE_U and RACE_P (it succeeds).
  5. Let the move finish.
a rotation move that meets same-named secrets created meanwhile is refused as a conflict
e2e/specs/secret-rotation/s8-rotation-move-race-answers-server-error.spec.ts:15
lines
infisical/backend/src/ee/services/secret-rotation-v2/secret-rotation-v2-service.ts:1134
infisical/backend/src/ee/services/secret-rotation-v2/secret-rotation-v2-service.ts:1186
infisical/backend/src/ee/services/secret-rotation-v2/secret-rotation-v2-service.ts:670
infisical/backend/src/server/plugins/error-handler.ts:274
before the PRs testedno
Assigning a machine identity to a project shows no options when the list fails to load
s6-failed-available-identities-load-shown-as-no-options-1
project-access-controluiloss: decision
medium

When You add an existing machine identity to a project and the list of identities can't be loaded.

Got The picker says "No options", as if your organisation had no identity left to add, and nothing tells you something failed.

Want The dialog says the identities couldn't be loaded and lets you retry.

steps
  1. Create a machine identity in the organisation without adding it to the project.
  2. Make the request listing the identities that can be assigned to the project fail with a server error.
  3. Open the project's Access Management page on the Machine Identities tab and click Add Machine Identity to Project.
  4. Choose Assign Existing and open the Machine Identity picker.
s6-failed-available-identities-load-shown-as-no-options
e2e/specs/project-access-control/s6-failed-available-identities-load-shown-as-no-options.spec.ts:12
lines
infisical/frontend/src/pages/project/AccessControlPage/components/IdentityTab/components/CreateProjectIdentity/CreateProjectIdentityForm.tsx:142
infisical/frontend/src/pages/project/AccessControlPage/components/IdentityTab/components/CreateProjectIdentity/CreateProjectIdentityForm.tsx:156
infisical/frontend/src/pages/project/AccessControlPage/components/IdentityTab/components/CreateProjectIdentity/CreateProjectIdentityForm.tsx:350
before the PRs testedno
Inviting many people at once does one database transaction per invitee
s13-bulk-invite-codes-written-per-invitee-1
org-membersperformanceloss: time
medium

When An admin invites many people to the organization in one go.

Got The invite does the same database work again for every person, all at once, so a large invite gets slower and can hold up everyone else using the app.

Want Inviting 50 people costs about the same database work as inviting 5.

steps
  1. Sign in as an organization admin
  2. Invite 5 new email addresses to the organization in one invite and count the database statements it runs
  3. Invite 50 new email addresses in one invite and count again
  4. Compare the statements on the auth_tokens table between the two invites
the sign-up codes of an invite are not written once per invitee
e2e/specs/org-members/s13-bulk-invite-codes-written-per-invitee.spec.ts:13
lines
infisical/backend/src/services/membership-user/org/org-membership-user-factory.ts:289
infisical/backend/src/services/auth-token/auth-token-service.ts:117
infisical/backend/src/services/auth-token/auth-token-service.ts:154
infisical/backend/src/server/routes/v1/invite-org-router.ts:29
before the PRs testedno
Listing machine identities to add to a project ignores the limit and returns all
s6-available-identities-ignore-limit-1
project-access-controlperformanceloss: time
medium

When You open the list of machine identities to add to a project, or ask the API for a page of them.

Got Every machine identity in the organisation comes back at once, whatever page size you ask for.

Want Only the page size asked for comes back.

steps
  1. Sign in as an organisation admin and create a new secret-management project.
  2. Create eight machine identities in the organisation without adding them to the project.
  3. Request the list of machine identities available to add to the project, with limit=5 (GET /api/v1/projects/<projectId>/memberships/available-identities?limit=5).
available identities for a project are bounded by the limit asked for
e2e/specs/project-access-control/s6-available-identities-ignore-limit.spec.ts:17
lines
infisical/backend/src/server/routes/v1/identity-project-membership-router.ts:612
infisical/backend/src/services/membership-identity/membership-identity-service.ts:577
infisical/backend/src/services/membership-identity/membership-identity-dal.ts:453
before the PRs testedno
Opening a project's Users tab loads every project member at once
s6-project-users-tab-loads-every-member-1
project-access-controlperformanceloss: time
medium

When Your project has many members and you open its Users tab.

Got The tab loads every member at once though it shows 20 a page, so it gets slower with every member added.

Want The tab loads only the page of members you are looking at.

steps
  1. Sign in as the admin of a project
  2. Add more than 20 organisation members to the project
  3. Open the project's Access Management, Users tab
  4. Compare the member list the tab loads with the rows the table shows
s6-project-users-tab-loads-every-member
e2e/specs/project-access-control/s6-project-users-tab-loads-every-member.spec.ts:15
lines
infisical/backend/src/server/routes/v1/project-router.ts:140
infisical/backend/src/server/routes/v1/project-router.ts:204
infisical/backend/src/services/project-membership/project-membership-service.ts:209
infisical/backend/src/services/project-membership/project-membership-dal.ts:27
infisical/frontend/src/pages/project/AccessControlPage/components/MembersTab/components/MembersTable.tsx:138
infisical/frontend/src/pages/project/AccessControlPage/components/MembersTab/components/MembersTable.tsx:209
before the PRs testedno
Listing a project's machine identities reads each identity's lockouts separately
s7-identity-list-reads-lockouts-per-identity-1
project-access-controlperformanceloss: time
low

When You open a project's machine identities, or the form to add one.

Got The product checks each identity's sign-in lockout separately, so the work grows with every identity listed.

Want The lockouts of all listed identities are checked together.

steps
  1. Create a secret-management project and attach one machine identity to it
  2. Watch the cache's command stream and list the project's machine identities
  3. Attach ten more machine identities and list them again while watching the command stream
s7-identity-list-reads-lockouts-per-identity: the project's identity list reads lockouts the same with 1 or 11 identities
e2e/specs/project-access-control/s7-identity-list-reads-lockouts-per-identity.spec.ts:31
lines
infisical/backend/src/services/membership-identity/membership-identity-service.ts:542
infisical/backend/src/services/membership-identity/membership-identity-service.ts:547
infisical/backend/src/services/identity/identity-fns.ts:38
before the PRs testedno
Opening a project's Groups tab loads every project group at once
s8-project-groups-tab-loads-every-group-1
project-access-controlperformanceloss: time
medium

When Your project has many groups and you open its Groups tab.

Got The tab loads every group at once though it shows 20 a page, so it gets slower with every group added.

Want The tab loads only the page of groups you are looking at.

steps
  1. Sign in as the admin of a project
  2. Create 40 organisation groups and add each to the project with the Viewer role
  3. Open the project's Access Management, Groups tab
  4. Compare the group list the tab loads with the rows the table shows
s8-project-groups-tab-loads-every-group
e2e/specs/project-access-control/s8-project-groups-tab-loads-every-group.spec.ts:15
lines
infisical/backend/src/server/routes/v1/project-group-memberships-router.ts:73
infisical/backend/src/services/membership-group/membership-group-service.ts:559
infisical/backend/src/services/membership-group/membership-group-dal.ts:216
infisical/frontend/src/pages/project/AccessControlPage/components/GroupsTab/components/GroupsSection/GroupsTable.tsx:91
infisical/frontend/src/pages/project/AccessControlPage/components/GroupsTab/components/GroupsSection/GroupsTable.tsx:118
before the PRs testedno
One connection of an unrecognised app blanks the whole App Connections tab
s6-unknown-app-connection-breaks-tab-1
app-connectionsbehaviourloss: time
low

When Your organisation has a connection of an app this version of the page does not recognise.

Got The App Connections tab shows only "The page broke" and none of your connections, even the ones it knows.

Want The tab lists the connections it recognises and leaves the odd one out or labels it generically.

steps
  1. Sign in as an organisation admin whose organisation has app connections (the shard has 20 AWS and 1 PostgreSQL connection).
  2. Have the app connections list name one connection's app as one the page does not know (here the PostgreSQL connection as "postgres-v2"; everything else unchanged).
  3. Open Integrations, App Connections tab.
s6-unknown-app-connection-breaks-tab
e2e/specs/app-connections/s6-unknown-app-connection-breaks-tab.spec.ts:26
lines
infisical/frontend/src/pages/organization/AppConnections/AppConnectionsPage/components/AppConnectionsTable.tsx:181
infisical/frontend/src/pages/organization/AppConnections/AppConnectionsPage/components/AppConnectionsTable.tsx:210
infisical/frontend/src/pages/organization/AppConnections/AppConnectionsPage/components/AppConnectionsTable.tsx:399
infisical/frontend/src/pages/organization/AppConnections/AppConnectionsPage/components/AppConnectionRow.tsx:105
before the PRs testedno
A failed folder create erases the folder name the user typed in the Add Folder dialog
s4-failed-folder-create-wipes-typed-name-1
behaviourloss: time
low

When

Got

Want

steps
  1. Open a project's dev secrets view
  2. Choose Add New, then Add Folder, and type a folder name
  3. Submit while the server fails the folder create
lines
infisical/frontend/src/pages/secret-manager/SecretDashboardPage/components/ActionBar/FolderForm.tsx:89
infisical/frontend/src/pages/secret-manager/OverviewPage/OverviewPage.tsx:1504
before the PRs testedno
Folder access given during a member's removal survives and returns on re-add
s9-folder-grant-survives-member-removal-1
project-access-controlbehaviourloss: decision
medium

When One admin gives a member folder access while another admin removes that member from the project at the same moment.

Got Both succeed, and the folder access quietly stays behind. If the person is added back later, the old folder access is in force again.

Want Folder access never outlives the membership: the grant is refused or removed with the member.

steps
  1. Have a project with a member holding the Viewer role.
  2. As a project admin, start granting that member Read folder access on the dev environment's root folder.
  3. Before the grant is saved, have a second admin remove the member from the project.
  4. Let both finish: both report success.
  5. Add the same person back to the project and open their folder access.
a folder grant arriving while the member is removed from the project leaves no grant behind
e2e/specs/project-access-control/s9-folder-grant-survives-member-removal.spec.ts:15
lines
infisical/backend/src/services/folder-permission/folder-permission-service.ts:104
infisical/backend/src/services/folder-permission/folder-permission-service.ts:113
infisical/backend/src/services/folder-permission/folder-permission-dal.ts:411
infisical/backend/src/services/membership-user/membership-user-service.ts:623
infisical/backend/src/services/membership-user/membership-user-service.ts:642
before the PRs testedno
A failed folder rename closes the Edit Folder dialog and discards the new name
s5-failed-folder-rename-loses-typed-name-1
behaviourloss: time
low

When

Got

Want

steps
  1. Open a project's dev secrets view at a folder holding a subfolder
  2. Open the subfolder's menu and choose Edit folder
  3. Type a new folder name
  4. Click Update Folder while the server fails the rename (500)
  5. The error message "Failed to rename folder across environments" shows and the Edit Folder dialog is gone, along with the typed name
lines
infisical/frontend/src/pages/secret-manager/OverviewPage/OverviewPage.tsx:1631
infisical/frontend/src/pages/secret-manager/OverviewPage/OverviewPage.tsx:1512
before the PRs testedno
A failed load of the organization roles says the organization has no roles
s1-org-roles-load-failure-shown-as-empty-1
behaviourloss: decision
medium

When

Got

Want

steps
  1. Sign in as an organization admin whose organization has its default roles (Admin, Member, No Access).
  2. Make the organization roles request fail with a server error.
  3. Open Organization Access Control and select the Roles tab.
  4. Wait for the load and its single automatic retry to fail.
lines
infisical/frontend/src/pages/organization/AccessManagementPage/components/OrgRoleTabSection/OrgRoleTable.tsx:93
infisical/frontend/src/pages/organization/AccessManagementPage/components/OrgRoleTabSection/OrgRoleTable.tsx:240
infisical/frontend/src/hooks/api/roles/queries.tsx:72
before the PRs testedno
A project's Roles tab says the project has no roles when its roles fail to load
s2-project-roles-load-failure-shown-as-empty-1
behaviourloss: decision
medium

When

Got

Want

steps
  1. Sign in as an organization admin and create a secret-management project (it gets its default roles).
  2. Make the project roles request fail with a server error (GET /api/v1/projects/<projectId>/roles answers 500).
  3. Open the project's Access Control and select the Roles tab.
  4. Wait for the read to give up (it is retried once).
lines
infisical/frontend/src/pages/project/AccessControlPage/components/ProjectRoleListTab/components/ProjectRoleList/ProjectRoleList.tsx:90
infisical/frontend/src/pages/project/AccessControlPage/components/ProjectRoleListTab/components/ProjectRoleList/ProjectRoleList.tsx:221
infisical/frontend/src/hooks/api/roles/queries.tsx:42
before the PRs testedno
Two role saves for one project member at once leave the member with both roles
s10-simultaneous-project-role-saves-keep-both-roles-1
project-access-controlbehaviourloss: decision
medium

When Two admins change the same project member's roles at the same moment, one to Member and one to Viewer.

Got Both saves succeed, and the member ends up holding both roles.

Want The member ends with the roles of one save, the one finished last.

steps
  1. Sign in as a project admin of a secret-management project and add two new members with the Viewer role.
  2. Save the first member's roles as Member, then as Viewer, one after the other: the member holds only Viewer.
  3. Send two role saves for the second member at the same moment: one setting the roles to [Member], one setting them to [Viewer].
  4. Both saves report success.
  5. Read the second member's project roles.
two simultaneous role saves of one project member leave exactly one project role
e2e/specs/project-access-control/s10-simultaneous-project-role-saves-keep-both-roles.spec.ts:13
lines
infisical/backend/src/services/membership-user/membership-user-service.ts:557
infisical/backend/src/services/membership-user/membership-user-service.ts:563
infisical/backend/src/services/membership-user/membership-user-service.ts:509
infisical/backend/src/db/migrations/20251005152640_simplify-membership.ts:130
infisical/frontend/src/pages/project/MemberDetailsByIDPage/components/MemberRoleDetailsSection/MemberMultiRoleModify.tsx:74
before the PRs testedno
A deleted organization role's page is left blank, with no not-found message and no back link
s3-deleted-org-role-page-blank-1
uiloss: time
medium

When

Got

Want

steps
  1. Sign in as an organization admin and create a custom organization role under Access Control > Roles
  2. Open the role's page and keep its link (/organizations/<orgId>/roles/<roleId>)
  3. Delete the role
  4. Open the kept link again
  5. Observe the content area: empty — no role title, no back link to Roles, no error or not-found message
lines
infisical/frontend/src/pages/organization/RoleByIDPage/RoleByIDPage.tsx:35
infisical/frontend/src/pages/organization/RoleByIDPage/RoleByIDPage.tsx:71
infisical/frontend/src/pages/organization/RoleByIDPage/route.tsx:5
before the PRs testedno
One connection with an unknown sign-in method breaks the App Connections search
s8-unknown-connection-method-breaks-tab-1
app-connectionsbehaviourloss: time
low

When One of your connections uses a sign-in method this version of the page does not recognise.

Got Searching the App Connections tab replaces it with "The page broke" and shows none of your connections.

Want The search lists the connection, perhaps without its method label.

steps
  1. Sign in as an organisation admin whose organisation has app connections (the shard has 20 AWS and 1 PostgreSQL connection).
  2. Have the app connections list give the PostgreSQL connection a sign-in method the page does not know (here "iam-auth"; everything else unchanged).
  3. Open Integrations, App Connections tab.
  4. Type the PostgreSQL connection's name into the search box.
s8-unknown-connection-method-breaks-tab
e2e/specs/app-connections/s8-unknown-connection-method-breaks-tab.spec.ts:26
lines
infisical/frontend/src/helpers/appConnections.ts:834
infisical/frontend/src/pages/organization/AppConnections/AppConnectionsPage/components/AppConnectionsTable.tsx:182
infisical/frontend/src/pages/organization/AppConnections/AppConnectionsPage/components/AppConnectionsTable.tsx:196
infisical/frontend/src/pages/organization/AppConnections/AppConnectionsPage/components/AppConnectionRow.tsx:104
before the PRs testedno
A project role page for a deleted or unknown role shows an empty page with no message
s4-deleted-project-role-page-blank-1
uiloss: time
medium

When

Got

Want

steps
  1. Sign in as an organization admin and open a secret-management project
  2. In Access Control > Roles, create a custom project role and copy the link to its page
  3. Delete that role
  4. Open the copied link
lines
infisical/frontend/src/pages/project/RoleDetailsBySlugPage/RoleDetailsBySlugPage.tsx:57
infisical/frontend/src/pages/project/RoleDetailsBySlugPage/RoleDetailsBySlugPage.tsx:101
infisical/frontend/src/pages/project/RoleDetailsBySlugPage/route-secret-manager.tsx:7
infisical/backend/src/services/role/role-service.ts:300
before the PRs testedno
Saving an open role page re-grants a permission another admin just revoked
s8-stale-role-page-save-regrants-revoked-policy-1
roles-permissionsbehaviourloss: work
medium

When You have a role's page open while another admin removes a permission from that role.

Got You add a different permission and save. The permission the other admin removed is back, and nothing tells either of you.

Want Their removal stays, or you are told the role changed since you opened it.

steps
  1. As an org admin, create a custom organization role granting Users read and Identities read.
  2. Open the role's page at Organization Access Control > Roles > the role; the page shows the Users policy.
  3. As another admin of the same organization, remove the Users read policy from the role and save (accepted).
  4. In the still-open role page, add an Audit Logs policy with Read and click [Save]; it says the role was updated.
  5. Read the role's permissions.
s8-stale-role-page-save-regrants-revoked-policy
e2e/specs/roles-permissions/s8-stale-role-page-save-regrants-revoked-policy.spec.ts:38
lines
infisical/frontend/src/pages/organization/RoleByIDPage/components/RolePermissionsSection/RolePermissionsSection.tsx:52
infisical/frontend/src/pages/organization/RoleByIDPage/components/RolePermissionsSection/RolePermissionsSection.tsx:67
infisical/backend/src/services/role/role-service.ts:159
infisical/frontend/src/hooks/api/reactQuery.tsx:349
before the PRs testedno
App Connections tab's view toggle and description fail accessibility checks
s7-connections-tab-a11y-violations-1
app-connectionsuiloss: nothing
low

When You use a screen reader or need strong contrast on the App Connections tab.

Got The Organization/All switch announces it controls a section that does not exist, and the grey description text under the title is slightly too faint.

Want The switch only refers to what is on the page and the description is easy to read.

steps
  1. Sign in as an organisation admin.
  2. Open Integrations, App Connections tab, and wait until the Add Connection button, the Organization/All toggle and the description under the title are shown.
  3. Run an accessibility check (axe) on the tab's content, outside the sidebar and header.
lines
infisical/frontend/src/pages/organization/AppConnections/AppConnectionsPage/components/AppConnectionsTable.tsx:341
infisical/frontend/src/pages/organization/AppConnections/AppConnectionsPage/components/AppConnectionsTable.tsx:349
infisical/frontend/src/components/v3/generic/Card/Card.tsx:49
infisical/frontend/src/pages/organization/AppConnections/AppConnectionsPage/components/AppConnectionsTable.tsx:304
before the PRs testedno
Listing an organisation's roles reads every organisation's roles
s8-org-role-list-reads-every-org-1
roles-permissionsperformanceloss: time
medium

When You open your organisation's roles, or a project's roles.

Got The product looks through every organisation's and project's roles on the instance to find yours, so the list slows down as the instance grows.

Want It finds only your organisation's or project's roles directly.

steps
  1. Sign in as an organisation admin, on an instance where other organisations and projects have their own custom roles
  2. Open Access Management and the Organization Roles tab (the org roles list)
  3. Read the indexes of the roles table in Postgres, and EXPLAIN the list statement
s8-org-role-list-reads-every-org
e2e/specs/roles-permissions/s8-org-role-list-reads-every-org.spec.ts:12
lines
infisical/backend/src/db/migrations/20251005152640_simplify-membership.ts:86
infisical/backend/src/db/migrations/20251005152640_simplify-membership.ts:108
infisical/backend/src/services/role/role-service.ts:214
infisical/backend/src/db/migrations/20260720120000_add-permission-query-fk-indexes.ts:16
before the PRs testedno
Deleting a custom role gets slower as the whole instance grows
s8-role-delete-scans-all-role-assignments-1
roles-permissionsperformanceloss: time
medium

When An admin deletes a custom role in an organization or a project.

Got To check nobody still holds the role, the product looks through every role assignment of every organization, so the delete slows as the instance grows.

Want Only that role's assignments are looked at, so the delete takes the same time however large the instance is.

steps
  1. Sign in as an organization admin
  2. Create a custom organization role that no member holds
  3. Delete the role from Access Management's Roles tab
  4. Read the database's index list for the role-assignments table (membership_roles)
s8-role-delete-scans-all-role-assignments
e2e/specs/roles-permissions/s8-role-delete-scans-all-role-assignments.spec.ts:10
lines
infisical/backend/src/services/role/role-service.ts:181
infisical/backend/src/db/migrations/20251005152640_simplify-membership.ts:139
infisical/backend/src/db/migrations/20251005152640_simplify-membership.ts:144
before the PRs testedno
Opening any folder gets slower with every folder in the environment
s16-folder-path-lookup-reads-every-folder-1
folders-imports-tagsperformanceloss: time
medium

When An environment holds many folders and you open or list any folder in it.

Got Each folder, import or secret request looks through every folder of the environment, so it gets slower as folders are added.

Want Opening a folder costs the same however many other folders the environment has.

steps
  1. Create a project; in its dev and staging environments add a folder /target with three subfolders
  2. In staging add 300 more folders at the root
  3. List the folders under /target in each environment (the dashboard does this when the folder is opened) and count the rows the database returns for the folder-path lookup
lines
infisical/backend/src/services/secret-folder/secret-folder-dal.ts:47
infisical/backend/src/services/secret-folder/secret-folder-dal.ts:146
infisical/backend/src/services/secret-folder/secret-folder-dal.ts:239
infisical/backend/src/services/secret-folder/secret-folder-service.ts:903
infisical/backend/src/db/migrations/20250319134021_recursive-folder-index.ts:10
before the PRs testedno
Opening a project's Roles tab loads every role at once
s9-project-roles-tab-loads-every-role-1
roles-permissionsperformanceloss: time
low

When Your project has many custom roles and you open its Roles tab.

Got The tab loads every role at once though it shows 20 a page, so it gets heavier with every role added.

Want The tab loads only the page of roles you are looking at.

steps
  1. Sign in as the admin of a secret-management project
  2. Create 30 custom project roles in it
  3. Open the project's Access Management, Roles tab
  4. Compare the role list the tab loads with the rows the table shows
s9-project-roles-tab-loads-every-role
e2e/specs/roles-permissions/s9-project-roles-tab-loads-every-role.spec.ts:12
lines
infisical/backend/src/ee/routes/v1/project-role-router.ts:298
infisical/backend/src/ee/routes/v1/org-role-router.ts:345
infisical/backend/src/services/role/role-service.ts:214
infisical/frontend/src/pages/project/AccessControlPage/components/ProjectRoleListTab/components/ProjectRoleList/ProjectRoleList.tsx:167
infisical/frontend/src/pages/organization/AccessManagementPage/components/OrgRoleTabSection/OrgRoleTable.tsx:196
before the PRs testedno
A member's access check reads rows multiplied by their roles and metadata
s10-permission-read-multiplies-roles-by-metadata-1
roles-permissionsperformanceloss: time
low

When A project member holds several roles and their organization profile carries metadata entries.

Got Every time their access is worked out, the product reads every role combined with every metadata entry, so the work multiplies as admins add either.

Want The work grows only by one per role and one per metadata entry.

steps
  1. In a secret-management project, create four custom project roles.
  2. Create a machine identity with ten metadata entries and add it to the project with the viewer, member and the four custom roles.
  3. Sign the identity in with universal auth and make its first request in the project (open the project).
  4. Read pg_stat_statements for the project permission statement (getPermission) and the permission fingerprint statement before and after that request.
A member's project permissions are read in rows that add up their roles and metadata, not multiply them
e2e/specs/roles-permissions/s10-permission-read-multiplies-roles-by-metadata.spec.ts:45
lines
infisical/backend/src/ee/services/permission/permission-dal.ts:226
infisical/backend/src/ee/services/permission/permission-dal.ts:251
infisical/backend/src/ee/services/permission/permission-dal.ts:978
infisical/backend/src/ee/services/permission/permission-dal.ts:994
infisical/backend/src/ee/services/permission/permission-service.ts:571
before the PRs testedno
A group's detail page that fails to load is left blank, with no error and no way back
s7-group-detail-failure-shown-blank-1
uiloss: time
medium

When

Got

Want

steps
  1. As an organisation admin, create a group in Access Management
  2. Open the group's detail page (Organization > Access Management > Groups > the group)
  3. Make the group details request fail with a server error (500) and reload the page
  4. Look at the page: the main area is empty
lines
infisical/frontend/src/pages/organization/GroupDetailsByIDPage/GroupDetailsByIDPage.tsx:43
infisical/frontend/src/pages/organization/GroupDetailsByIDPage/GroupDetailsByIDPage.tsx:87
infisical/frontend/src/hooks/api/groups/queries.tsx:108
before the PRs testedno
A group's project page that fails to load says "Group Not Found ... no longer assigned to the project"
s5-group-project-detail-failure-shown-as-not-found-1
uiloss: decision
medium

When

Got

Want

steps
  1. As an organisation admin, create a group and add it to a project as Viewer
  2. Open the group's detail page inside the project (Project > Access Control > Groups > the group)
  3. Make the group's project membership request fail with a server error (500) and reload the page
  4. Read the page: it says the group was not found or is no longer assigned to the project
lines
infisical/frontend/src/pages/project/GroupDetailsByIDPage/GroupDetailsByIDPage.tsx:47
infisical/frontend/src/pages/project/GroupDetailsByIDPage/GroupDetailsByIDPage.tsx:203
infisical/frontend/src/hooks/api/projects/queries.tsx:542
before the PRs testedno
Searching a group's members by full name, or its projects by a name with a space, says nothing matches
s4-group-search-with-space-finds-nothing-1
behaviourloss: time
medium

When

Got

Want

steps
  1. Sign in as an organisation admin whose profile has a first and last name set.
  2. Create a group under Access Management > Groups and add yourself as a member.
  3. Create a project whose name contains a space (e.g. "Billing Alpha") and give the group Viewer access to it.
  4. Open the group's detail page: the member and the project are both listed.
  5. Type the member's "First Last" in Search members... — the table says "No members match this search".
  6. Type the project's full name in Search projects... — the table says "No projects match this search".
lines
infisical/backend/src/ee/routes/v1/group-router.ts:436
infisical/backend/src/ee/routes/v1/group-router.ts:530
infisical/backend/src/ee/services/group/group-dal.ts:397
infisical/backend/src/ee/services/group/group-dal.ts:524
infisical/backend/src/ee/routes/v1/group-router.ts:323
infisical/frontend/src/pages/organization/GroupDetailsByIDPage/components/GroupMembersSection/GroupMembersTable.tsx:133
infisical/frontend/src/pages/organization/GroupDetailsByIDPage/components/GroupProjectsSection/GroupProjectsTable.tsx:113
before the PRs testedno
A group's project search finds nothing when the search has a space
s9-group-project-search-with-space-finds-nothing-1
groupsbehaviourloss: time
medium

When You search a group's projects for two words of a project's name, like 'Alpha Project'.

Got The group's Projects section says no projects match, though the project is right there without the search.

Want The project whose name contains the words is listed.

steps
  1. Sign in as an organisation admin and create a group
  2. Create a project named '<anything> Alpha Project' and add the group to it as Viewer
  3. Open Organization › Access Control › Groups and open the group's detail page; the project is listed under Projects
  4. Type 'Alpha Project' in the Projects section's 'Search projects...' box
a group's project is found in its Projects search by two words of its name
e2e/specs/groups/s9-group-project-search-with-space-finds-nothing.spec.ts:9
lines
infisical/backend/src/ee/routes/v1/group-router.ts:527
infisical/backend/src/ee/services/group/group-dal.ts:523
infisical/frontend/src/pages/organization/GroupDetailsByIDPage/components/GroupProjectsSection/GroupProjectsTable.tsx:78
infisical/frontend/src/pages/organization/GroupDetailsByIDPage/components/AddGroupProjectModal.tsx:59
before the PRs testedno
A group whose members or projects fail to load is shown as having no members / no projects
s1-group-members-failure-shown-as-empty-1
behaviourloss: decision
medium

When

Got

Want

steps
  1. As an organisation admin, create a group, add yourself to it, and add the group to a project as Viewer
  2. Make the group's member list request fail with a server error
  3. Open the group's page under Access Management (/organizations/:orgId/groups/:groupId)
  4. The Group Members section says "This group does not have any members" with an Add Members button, and no error
  5. Do the same with the group's project list request failing: the Projects section says "This group is not a member of any projects"
  6. Open the same group inside the project (/organizations/:orgId/projects/secret-management/:projectId/groups/:groupId) with the member list failing: it says "This group doesn't have any members"
lines
infisical/frontend/src/pages/organization/GroupDetailsByIDPage/components/GroupMembersSection/GroupMembersTable.tsx:87
infisical/frontend/src/pages/organization/GroupDetailsByIDPage/components/GroupMembersSection/GroupMembersTable.tsx:214
infisical/frontend/src/pages/organization/GroupDetailsByIDPage/components/GroupProjectsSection/GroupProjectsTable.tsx:78
infisical/frontend/src/pages/organization/GroupDetailsByIDPage/components/GroupProjectsSection/GroupProjectsTable.tsx:149
infisical/frontend/src/pages/project/GroupDetailsByIDPage/components/GroupMembersSection/GroupMembersTable.tsx:98
infisical/frontend/src/pages/project/GroupDetailsByIDPage/components/GroupMembersSection/GroupMembersTable.tsx:262
infisical/frontend/src/hooks/api/groups/queries.tsx:168
infisical/frontend/src/hooks/api/groups/queries.tsx:271
before the PRs testedno
Signing in with LDAP twice at once adds the user to their mapped group twice
s21-simultaneous-ldap-sign-ins-duplicate-group-membership-1
sso-loginbehaviourloss: nothing
low

When You sign in with LDAP from two tabs at the same moment.

Got You end up in your mapped group twice.

Want You are in the group once.

steps
  1. Enable LDAP for the organization with a group search base, and map a directory group to a new Infisical group
  2. As a directory user in that directory group who has signed in before, sign in with LDAP twice at the same moment
  3. Look at the Infisical group's memberships for that user
lines
infisical/backend/src/ee/services/ldap-config/ldap-config-service.ts:683
infisical/backend/src/ee/services/ldap-config/ldap-config-service.ts:697
infisical/backend/src/ee/services/group/group-fns.ts:214
infisical/backend/src/ee/services/group/group-fns.ts:297
infisical/backend/src/db/migrations/20260728120000_add-user-group-membership-indexes.ts:30
infisical/backend/src/ee/services/scim/scim-service.ts:1646
infisical/backend/src/ee/services/group/group-service.ts:1090
before the PRs testedno
A sync link with an unknown destination crashes the page
s5-unknown-destination-link-crashes-1
secret-syncsuiloss: nothing
low

When You open a sync link whose destination part is mistyped or unknown.

Got The whole page turns into an error screen with a Retry button.

Want The page says it could not find that sync.

steps
  1. Sign in and open a project in Secrets Management
  2. Open /organizations/<orgId>/projects/secret-management/<projectId>/integrations/secret-syncs/not-a-destination/00000000-0000-0000-0000-000000000000
lines
infisical/frontend/src/pages/secret-manager/SecretSyncDetailsByIDPage/SecretSyncDetailsByIDPage.tsx:70
infisical/frontend/src/pages/secret-manager/SecretSyncDetailsByIDPage/route.tsx:8
before the PRs testedno
A folder grant survives when a user leaves both their groups at the same moment
s11-concurrent-group-removals-keep-folder-grant-1
groupsbehaviourloss: decision
medium

When A user is removed from two groups that both give access to the same project at the same moment.

Got The user is out of both groups, but the folder access you gave them on that project is still on record.

Want Leaving their last way into the project removes their folder access, as it does when the removals happen one after the other.

steps
  1. Sign up a user and invite them to the organisation as a member.
  2. Create groups A and B and add both to a project as Viewer.
  3. Add the user to A and B.
  4. Give the user read folder access on the project's dev environment root.
  5. Control: remove the user from A, then from B — the folder grant is gone after the second removal.
  6. Re-add the user to A and B and grant the folder access again.
  7. Hold the user's two group-membership rows so both removals pass their 'does the user still reach the project' check before either deletes its row, and remove the user from A and from B at the same moment.
  8. Count the user's additional privileges on the project.
removing a user from both groups that reach a project at once leaves no folder grant behind
e2e/specs/groups/s11-concurrent-group-removals-keep-folder-grant.spec.ts:14
lines
infisical/backend/src/ee/services/group/group-fns.ts:497
infisical/backend/src/ee/services/group/group-fns.ts:506
infisical/backend/src/ee/services/group/group-fns.ts:522
infisical/backend/src/ee/services/group/group-fns.ts:533
infisical/backend/src/ee/services/group/user-group-membership-dal.ts:39
before the PRs testedno
Two role changes to a group at once leave the group with both roles
s11-simultaneous-group-role-changes-leave-two-roles-1
groupsbehaviourloss: decision
medium

When Two admins change the same group's organisation role at the same moment.

Got Both saves succeed, and the group ends up holding both roles, so its members get the rights of both.

Want The group ends with one role, the one saved last.

steps
  1. Sign in as an organisation admin and create a group with the Member role.
  2. Send two role changes for that group at the same moment: one to Admin, one to No Access (two admins, or two tabs, saving the group's role).
  3. Both saves report success.
  4. Read the group's organisation roles.
two simultaneous role changes of one group leave exactly one org role
e2e/specs/groups/s11-simultaneous-group-role-changes-leave-two-roles.spec.ts:24
lines
infisical/backend/src/ee/services/group/group-service.ts:348
infisical/backend/src/ee/services/group/group-service.ts:359
infisical/backend/src/ee/services/group/group-service.ts:360
infisical/backend/src/db/migrations/20251005152640_simplify-membership.ts:144
before the PRs testedno
Adding someone to a group gets slower with every project the group can access
s11-group-member-add-reads-projects-one-by-one-1
groupsperformanceloss: time
low

When An admin adds someone to a group that has access to many projects.

Got The product looks up each of the group's projects one at a time, so the add takes longer the more projects the group can access.

Want Adding a member costs the same whether the group has access to 5 projects or 100.

steps
  1. As an organisation admin, create two groups and give one access to 6 projects and the other access to 96 projects (Project › Access Control › Groups › Add Group to Project).
  2. Add an accepted organisation member to the first group, then remove them; repeat three times while counting the database statements of each add.
  3. Do the same with the second group.
  4. Compare the number of project lookups each add made.
s11-group-member-add-reads-projects-one-by-one
e2e/specs/groups/s11-group-member-add-reads-projects-one-by-one.spec.ts:39
lines
infisical/backend/src/ee/services/group/group-fns.ts:75
infisical/backend/src/ee/services/group/group-fns.ts:76
infisical/backend/src/ee/services/group/group-fns.ts:63
infisical/backend/src/ee/services/group/group-fns.ts:283
infisical/backend/src/ee/routes/v1/group-router.ts:576
before the PRs testedno
A group's Projects list hides its Certificate Manager projects
s11-group-projects-hide-certificate-manager-1
groupsuiloss: decision
high

When You open a group that belongs to a Certificate Manager project.

Got That project is never listed. The pager still counts it, and a group in only such a project is described as in no projects.

Want Every project the group belongs to is listed, and the count matches the list.

steps
  1. As an org admin, create a group in the organisation.
  2. Add the group to a Certificate Manager project as Member.
  3. Optionally also add it to a secret-management project as Viewer.
  4. Open the group's detail page under Access Management > Groups and look at the Projects section.
a group's project count matches the projects its list shows
e2e/specs/groups/s11-group-projects-hide-certificate-manager.spec.ts:34
lines
infisical/frontend/src/pages/organization/GroupDetailsByIDPage/components/GroupProjectsSection/GroupProjectsTable.tsx:90
infisical/frontend/src/pages/organization/GroupDetailsByIDPage/components/GroupProjectsSection/GroupProjectsTable.tsx:178
infisical/frontend/src/pages/organization/GroupDetailsByIDPage/components/GroupProjectsSection/GroupProjectsTable.tsx:149
infisical/backend/src/ee/services/group/group-dal.ts:507
before the PRs testedno
A failed machine identity list load says the organization has no machine identities
s2-identity-list-failure-shown-as-empty-1
behaviourloss: decision
medium

When

Got

Want

steps
  1. Sign in as an organization admin.
  2. Make the machine identity search request fail with a server error.
  3. Open Organization Access Control and select the Machine Identities tab.
  4. Wait for the load and its single automatic retry to fail.
lines
infisical/frontend/src/pages/organization/AccessManagementPage/components/OrgIdentityTab/components/IdentitySection/IdentityTable.tsx:525
infisical/frontend/src/pages/organization/AccessManagementPage/components/OrgIdentityTab/components/IdentitySection/IdentityTable.tsx:567
infisical/frontend/src/pages/organization/AccessManagementPage/components/OrgIdentityTab/components/IdentitySection/IdentityTable.tsx:624
infisical/frontend/src/hooks/api/identities/queries.tsx:84
before the PRs testedno
A sub-organisation's group link made during the group's deletion is silently erased
s12-group-linked-during-delete-deleted-anyway-1
groupsbehaviourloss: work
medium

When A sub-organisation admin links a parent group just as a root admin deletes that group.

Got Both are told it worked. The group is gone and the sub-organisation's link disappears with it.

Want The delete is refused because the sub-organisation now links the group.

steps
  1. Sign in as a root-organisation admin on the enterprise plan and create a sub-organisation.
  2. Create a group in the root organisation, invite a new member into it, give the group Viewer access to a new project, and give that member a folder grant on the project.
  3. Control: link a second group to the sub-organisation as the sub-organisation's admin, then delete it from the root organisation: the delete is refused because a sub-organisation links it.
  4. Start deleting the first group from the root organisation and, while the delete is still removing the member's folder grant, link the same group to the sub-organisation as its admin.
  5. Let both finish and look at the sub-organisation's groups.
s12-group-linked-during-delete-deleted-anyway
e2e/specs/groups/s12-group-linked-during-delete-deleted-anyway.spec.ts:21
lines
infisical/backend/src/ee/services/group/group-service.ts:660
infisical/backend/src/ee/services/group/group-service.ts:663
infisical/backend/src/ee/services/group/group-service.ts:668
infisical/backend/src/ee/services/group/group-service.ts:682
infisical/backend/src/ee/services/group/group-folder-grant-fns.ts:165
infisical/backend/src/db/migrations/20251005152640_simplify-membership.ts:29
before the PRs testedno
The secrets page's toolbar is hard to use with a screen reader
s9-secrets-page-toolbar-a11y-violations-1
secret-syncsuiloss: nothing
low

When You open a project's secrets page with a screen reader or low vision.

Got The environment picker is announced as an extra unnamed button, the search box reports a state it cannot have, and the Name header is faint.

Want Each control is announced once with its name, and the header is readable.

steps
  1. Sign in and open a project's secrets page for an environment (here the shard project's dev environment)
  2. Wait for the secrets table to load
  3. Run an axe accessibility check on the page, leaving out the navigation shell and the page-loading bar
s9-secrets-page-toolbar-a11y-violations
e2e/specs/secret-syncs/s9-secrets-page-toolbar-a11y-violations.spec.ts:10
lines
infisical/frontend/src/pages/secret-manager/OverviewPage/components/EnvironmentSelect/EnvironmentSelect.tsx:121
infisical/frontend/src/pages/secret-manager/OverviewPage/components/ResourceSearchInput/ResourceSearchInput.tsx:86
infisical/frontend/src/pages/secret-manager/OverviewPage/OverviewPage.tsx:3273
before the PRs testedno
A machine identity's page says it is not a member of any projects when its project memberships fail to load
s4-identity-projects-failure-shown-as-not-member-1
behaviourloss: decision
medium

When

Got

Want

steps
  1. Sign in as an organization admin, create a machine identity and add it to a project as a viewer
  2. Make the request that lists the identity's project memberships fail with a server error
  3. Open Access Control > Machine Identities and open that identity's details page
  4. Look at the Projects section
lines
infisical/frontend/src/pages/organization/IdentityDetailsByIDPage/components/IdentityProjectsSection/IdentityProjectsTable.tsx:48
infisical/frontend/src/pages/organization/IdentityDetailsByIDPage/components/IdentityProjectsSection/IdentityProjectsTable.tsx:150
infisical/frontend/src/hooks/api/identities/queries.tsx:121
before the PRs testedno
A failed load of a project's machine identity page says the identity was not found or removed
s3-project-identity-failure-shown-as-not-found-1
behaviourloss: decision
medium

When

Got

Want

steps
  1. Sign in as an organization admin and add a machine identity to a project
  2. Make the server fail the identity's project membership read (it answers 500)
  3. Open the machine identity's page inside the project
  4. Wait for the read and its one retry to fail
  5. The page says 'Machine identity not found — The machine identity may have been removed or you may not have access to it.' and shows no error
lines
infisical/frontend/src/pages/project/IdentityDetailsByIDPage/IdentityDetailsByIDPage.tsx:90
infisical/frontend/src/pages/project/IdentityDetailsByIDPage/IdentityDetailsByIDPage.tsx:468
infisical/frontend/src/hooks/api/projectIdentityMembership/queries.ts:111
before the PRs testedno
A machine identity's details page is left blank, with no message and no back link, when the identity cannot be loaded
s3-identity-detail-failure-shown-blank-1
uiloss: time
medium

When

Got

Want

steps
  1. Sign in as an organization admin and create a machine identity
  2. Make the machine identity's load fail with a server error
  3. Open the identity's details page (/organizations/<orgId>/identities/<identityId>)
  4. Wait for the load and its single automatic retry to fail
  5. Observe the content area: empty — no identity name, no back link, no error message
lines
infisical/frontend/src/pages/organization/IdentityDetailsByIDPage/IdentityDetailsByIDPage.tsx:59
infisical/frontend/src/pages/organization/IdentityDetailsByIDPage/IdentityDetailsByIDPage.tsx:96
infisical/frontend/src/hooks/api/identities/queries.tsx:69
before the PRs testedno
Listing machine identities through the API returns all of them at once
s8-public-identity-list-unbounded-1
machine-identitiesperformanceloss: time
medium

When An organization with many machine identities lists them through the API.

Got Every identity comes back in one reply, and asking for fewer is ignored.

Want One page at a time, like the identity search.

steps
  1. Sign in as an organization admin whose organization holds more than 100 machine identities (create them through the product's Create Identity action or its API)
  2. List the organization's machine identities through the public API (listMachineIdentities, GET /api/v1/identities?orgId=<org>)
  3. Repeat with &limit=5 appended
  4. Count the identities in each reply
s8 the public machine identity list returns at most one page
e2e/specs/machine-identities/s8-public-identity-list-unbounded.spec.ts:11
lines
infisical/backend/src/server/routes/v1/identity-router.ts:329
infisical/backend/src/server/routes/v1/identity-router.ts:351
infisical/backend/src/services/identity/identity-service.ts:518
infisical/backend/src/services/identity/identity-org-dal.ts:187
before the PRs testedno
The Identities tab slows with every project a non-admin member belongs to
s9-identity-count-loads-permission-per-project-1
machine-identitiesperformanceloss: time
medium

When You open the Identities tab as a regular organization member, not an admin.

Got The tab does extra work for every project you belong to, each time it opens and each time you search.

Want The tab costs the same however many projects you are in.

steps
  1. Sign up a new user and invite them into the organization with the member role, then accept the invite
  2. Create two projects as the organization admin and add the member to each with the viewer role
  3. As the member, count the organization's and the projects' machine identities (what the Identities tab sends on open) and watch the permission-cache reads that name those projects
  4. Create ten more projects and add the member to each
  5. Count the machine identities again as the member and watch the same reads
s9-identity-count-loads-permission-per-project: the Identities tab's count reads the member's permissions the same with 2 or 12 projects
e2e/specs/machine-identities/s9-identity-count-loads-permission-per-project.spec.ts:32
lines
infisical/backend/src/services/identity-v2/identity-service.ts:533
infisical/backend/src/services/identity-v2/identity-service.ts:536
infisical/backend/src/ee/services/permission/permission-service.ts:633
infisical/frontend/src/pages/organization/AccessManagementPage/components/OrgIdentityTab/components/IdentitySection/IdentityTable.tsx:535
before the PRs testedno
A machine identity's page loads every project it belongs to at once
s8-identity-projects-loads-every-membership-1
machine-identitiesperformanceloss: time
medium

When A machine identity belongs to many projects and you open its page.

Got The page loads every project it belongs to at once though it shows 20 at a time, so it gets slower with each project added.

Want The page loads only the page of projects you are looking at.

steps
  1. Sign in as an organization admin
  2. Create a machine identity and add it to 30 projects
  3. Open the machine identity's detail page
  4. Compare the project memberships the page loads with the rows its Projects table shows
s8-identity-projects-loads-every-membership
e2e/specs/machine-identities/s8-identity-projects-loads-every-membership.spec.ts:11
lines
infisical/backend/src/server/routes/v1/identity-router.ts:466
infisical/backend/src/services/identity/identity-service.ts:604
infisical/backend/src/services/identity-project/identity-project-dal.ts:33
infisical/frontend/src/pages/organization/IdentityDetailsByIDPage/components/IdentityProjectsSection/IdentityProjectsTable.tsx:139
before the PRs testedno
Searching the organization's machine identities checks each one's lockouts separately
s8-org-identity-search-reads-lockouts-per-identity-1
machine-identitiesperformanceloss: time
low

When You open or search the organization's machine identities.

Got The product checks each identity's sign-in lockout separately, so the work grows with every identity on the page.

Want The lockouts of all listed identities are checked together.

steps
  1. Sign in as an organization admin and create five machine identities
  2. Watch the cache's command stream and search the organization's machine identities for those five
  3. Create forty-five more and search for all fifty while watching the command stream
s8-org-identity-search-reads-lockouts-per-identity: the organization's identity search reads lockouts the same with 5 or 50 identities
e2e/specs/machine-identities/s8-org-identity-search-reads-lockouts-per-identity.spec.ts:32
lines
infisical/backend/src/services/identity-v2/identity-service.ts:609
infisical/backend/src/services/identity-v2/identity-service.ts:614
infisical/backend/src/services/identity/identity-service.ts:567
infisical/backend/src/services/identity/identity-service.ts:572
infisical/backend/src/services/identity/identity-fns.ts:38
before the PRs testedno
Two updates of a machine identity at once leave it with both roles
s10-simultaneous-identity-updates-keep-both-sets-1
machine-identitiesbehaviourloss: decision
medium

When Two scripts or tools change the same machine identity's role or metadata at the same moment.

Got Both changes report success, and the identity ends up holding both roles, or both sets of metadata.

Want The identity keeps one role and one metadata set, the last one saved.

steps
  1. Sign in as an org admin and create a machine identity with the Member role.
  2. As a control, change another identity's role to Member, then to No Access, one after the other: it holds only No Access.
  3. Through the public API, send two updates of the first identity at the same moment, one with role Member and one with role No Access, neither with a name.
  4. Both updates report success.
  5. Read the identity's organization roles.
  6. Likewise send two metadata updates at the same moment, one with key a and one with key b, and read its metadata.
two simultaneous role changes of one machine identity leave exactly one org role
e2e/specs/machine-identities/s10-simultaneous-identity-updates-keep-both-sets.spec.ts:19
lines
infisical/backend/src/services/identity/identity-service.ts:294
infisical/backend/src/services/identity/identity-service.ts:300
infisical/backend/src/services/identity/identity-service.ts:317
infisical/backend/src/services/membership-identity/membership-identity-service.ts:325
before the PRs testedno
A late sign-up code fails with "Something went wrong" instead of a clear refusal
s11-late-signup-code-answers-server-error-1
project-access-controlbehaviourloss: time
low

When You confirm sign-up in one tab, ask for a new code in another, and finish your account in the first.

Got Entering the second code shows "Something went wrong".

Want Infisical says the account already exists and points you to sign in.

steps
  1. Start sign-up with a new email in tab one and enter the emailed code there.
  2. In tab two, start sign-up again with the same email and receive a second code.
  3. In tab one, finish the account (name, password, organisation).
  4. In tab two, enter the second code.
a sign-up code entered after the account is completed is refused, not a server error
e2e/specs/project-access-control/s11-late-signup-code-answers-server-error.spec.ts:10
lines
infisical/backend/src/services/auth/auth-signup-service.ts:224
infisical/backend/src/services/auth/email-dispatch-guard.ts:105
before the PRs testedno
Pressing Escape on a new project throws away the typed name without asking
s6-new-project-name-lost-on-escape-1
uiloss: time
low

When

Got

Want

steps
  1. Sign in and open the organisation's Secrets Management projects page
  2. Click New Project
  3. Type a project name and a description
  4. Press Escape (or click Cancel)
  5. Click New Project again
lines
infisical/frontend/src/components/projects/NewProjectModal.tsx:390
infisical/frontend/src/pages/organization/ProjectsPage/ProjectTypePage/ProjectTypePage.tsx:388
infisical/frontend/src/components/projects/NewProjectModal.tsx:108
infisical/frontend/src/components/projects/NewProjectModal.tsx:352
before the PRs testedno
A machine identity is deleted even though delete protection was just turned on
s11-identity-protected-during-delete-deleted-1
machine-identitiesbehaviourloss: work
medium

When One admin turns on delete protection for a machine identity while another deletes it at the same moment.

Got Both are told they succeeded, and the machine identity is deleted anyway.

Want The identity stays protected and the delete is refused, or the protection save says it is gone.

steps
  1. As an org admin, create a machine identity with delete protection off.
  2. As one admin, turn delete protection on for the identity and save.
  3. At the same moment, as another admin, delete the identity.
  4. Look for the identity in the organization's Machine Identities list.
a machine identity whose delete protection is turned on during its deletion is not deleted
e2e/specs/machine-identities/s11-identity-protected-during-delete-deleted.spec.ts:18
lines
infisical/backend/src/services/identity/identity-service.ts:375
infisical/backend/src/services/identity/identity-service.ts:419
infisical/backend/src/services/identity/identity-service.ts:423
infisical/backend/src/services/identity/identity-service.ts:432
infisical/backend/src/services/identity/identity-service.ts:296
before the PRs testedno
Cancelling Add Folder throws away the typed folder name without asking
s17-add-folder-draft-discarded-on-cancel-1
folders-imports-tagsuiloss: time
low

When You start adding a folder, type its name, then press Cancel.

Got The dialog closes straight away and the name you typed is gone when you open it again.

Want You are asked whether to discard what you typed, or the name is kept.

steps
  1. Sign in and open a secret-management project's dev environment secrets screen
  2. Choose Add New, then Add Folder
  3. Type a folder name
  4. Press Cancel
  5. Choose Add New, then Add Folder again
s17-add-folder-draft-discarded-on-cancel
e2e/specs/folders-imports-tags/s17-add-folder-draft-discarded-on-cancel.spec.ts:8
lines
infisical/frontend/src/pages/secret-manager/SecretDashboardPage/components/ActionBar/FolderForm.tsx:139
infisical/frontend/src/pages/secret-manager/OverviewPage/OverviewPage.tsx:4004
before the PRs testedno
Renaming a machine identity can undo another admin's delete protection and metadata
s12-stale-identity-edit-keeps-newer-changes-1
machine-identitiesbehaviourloss: work
low

When You rename a machine identity in a dialog you opened before another admin turned on its delete protection and added metadata.

Got Your save quietly turns delete protection back off and removes the metadata they added.

Want Only the name changes, or you are told the identity changed since you opened the dialog.

steps
  1. As an org admin, create a machine identity with delete protection off and one metadata entry k1.
  2. Open its detail page and open the Update Machine Identity dialog.
  3. As another admin (here: the API with the same org's token), turn delete protection on and set metadata to k1 and k2.
  4. Back in the still-open dialog, change only the name and click Update.
  5. Read the identity back.
renaming from an edit dialog opened before another admin's change keeps that change
e2e/specs/machine-identities/s12-stale-identity-edit-keeps-newer-changes.spec.ts:9
lines
infisical/frontend/src/pages/organization/AccessManagementPage/components/OrgIdentityTab/components/IdentitySection/OrgIdentityModal.tsx:103
infisical/frontend/src/pages/organization/AccessManagementPage/components/OrgIdentityTab/components/IdentitySection/OrgIdentityModal.tsx:129
infisical/backend/src/services/identity/identity-service.ts:296
infisical/backend/src/services/identity/identity-service.ts:317
before the PRs testedno
Deleting a role at the same moment as another admin shows "Something went wrong"
s11-concurrent-org-role-delete-server-error-1
roles-permissionsbehaviourloss: time
low

When You delete a custom organization role at the same moment as another admin does.

Got The role is deleted, but one of you is told "Something went wrong".

Want The second admin is told the role no longer exists.

steps
  1. Sign in as an organization admin and create a custom organization role.
  2. Send two deletes of that role so both reach the server at the same moment (the test holds the role's row so both pass the existence check, then lets them go).
  3. Read both answers.
s11-concurrent-org-role-delete-server-error
e2e/specs/roles-permissions/s11-concurrent-org-role-delete-server-error.spec.ts:9
lines
infisical/backend/src/services/role/role-service.ts:175
infisical/backend/src/services/role/role-service.ts:198
infisical/backend/src/services/role/role-service.ts:203
infisical/backend/src/ee/routes/v1/org-role-router.ts:280
before the PRs testedno
A machine identity with more than 20 access tokens shows only 20 of them, and the rest can never be reached or revoked from the screen
s2-token-auth-list-capped-at-20-1
behaviourloss: decision
high

When

Got

Want

steps
  1. Create a machine identity in an organization and add Token Auth to it
  2. Create 21 access tokens on it with Add Token
  3. Open the identity's page and click Token Auth in the Authentication card
  4. Read the Access Tokens pager and page to the end
lines
infisical/frontend/src/hooks/api/identities/queries.tsx:373
infisical/frontend/src/views/IdentityAuthMethods/helpers/IdentityTokenAuthTokensTable.tsx:136
infisical/frontend/src/views/IdentityAuthMethods/helpers/IdentityTokenAuthTokensTable.tsx:254
infisical/backend/src/server/routes/v1/identity-token-auth-router.ts:447
infisical/backend/src/services/identity-token-auth/identity-token-auth-service.ts:611
before the PRs testedno
A failed load of an identity's access tokens says none have been generated
s5-access-token-list-failure-shown-as-empty-1
identity-authbehaviourloss: decision
medium

When You open a machine identity's Token Auth method and its access tokens cannot be loaded.

Got It says no access tokens have been generated, with no error, so live tokens look absent and cannot be revoked from the screen.

Want It says the tokens could not be loaded and lets you retry.

steps
  1. Sign in as an organisation admin.
  2. Create a machine identity, add Token Auth to it and create one access token.
  3. Make the access token list read fail with a server error.
  4. Open the identity's page and click Token Auth in its Authentication card.
  5. Wait for the load and its single automatic retry to fail.
s5-access-token-list-failure-shown-as-empty
e2e/specs/identity-auth/s5-access-token-list-failure-shown-as-empty.spec.ts:13
lines
infisical/frontend/src/views/IdentityAuthMethods/content/IdentityTokenAuthContent.tsx:18
infisical/frontend/src/views/IdentityAuthMethods/helpers/IdentityTokenAuthTokensTable.tsx:123
infisical/frontend/src/views/IdentityAuthMethods/helpers/IdentityTokenAuthTokensTable.tsx:272
infisical/frontend/src/hooks/api/identities/queries.tsx:364
before the PRs testedno
A newly created access token is announced as an updated token
s5-new-access-token-confirmed-as-created-1
identity-authuiloss: nothing
low

When You add a new access token to a machine identity.

Got The confirmation says the token was updated.

Want It says the token was created.

steps
  1. Sign in as an org admin and open a machine identity that has Token Auth
  2. Click the Token Auth row in the Authentication card
  3. Click Add Token, enter a name, click Create
  4. Read the success notification
s5-new-access-token-confirmed-as-created
e2e/specs/identity-auth/s5-new-access-token-confirmed-as-created.spec.ts:9
lines
infisical/frontend/src/pages/organization/IdentityDetailsByIDPage/components/IdentityTokenModal.tsx:104
infisical/frontend/src/views/IdentityAuthMethods/helpers/IdentityTokenAuthTokensTable.tsx:112
before the PRs testedno
An identity's client secrets look like they were never created when they fail to load
s5-client-secret-list-load-error-shown-as-none-1
identity-authbehaviourloss: decision
medium

When You open a machine identity's Universal Auth method and its client secrets cannot be loaded.

Got It says no client secrets have been generated, with no error, though one exists and still works.

Want It tells you the client secrets could not be loaded.

steps
  1. Sign in as an organization admin, create a machine identity, add Universal Auth and create one client secret
  2. Make the request that lists the identity's client secrets fail with a server error
  3. Open Access Control > Machine Identities, open that identity and open its Universal Auth method from the Authentication section
  4. Look at the Client Secrets section
client secret list that failed to load does not say none were generated
e2e/specs/identity-auth/s5-client-secret-list-load-error-shown-as-none.spec.ts:12
lines
infisical/frontend/src/views/IdentityAuthMethods/content/IdentityUniversalAuthContent.tsx:36
infisical/frontend/src/views/IdentityAuthMethods/helpers/IdentityUniversalAuthClientSecretsTable.tsx:115
infisical/frontend/src/views/IdentityAuthMethods/helpers/IdentityUniversalAuthClientSecretsTable.tsx:212
infisical/frontend/src/hooks/api/identities/queries.tsx:160
before the PRs testedno
Machine logins look through every identity's universal auth to find one client ID
s5-universal-auth-login-scans-client-ids-1
identity-authperformanceloss: time
medium

When A machine (a CI job, a service, an agent) logs in with Universal Auth.

Got The product looks through the login settings of every machine identity on the instance to find yours, so logins get slower as identities are added anywhere.

Want Your client ID is found directly, at the same cost however many identities exist.

steps
  1. In an organisation, create a machine identity and add Universal Auth to it
  2. Create a client secret for it
  3. Log in as the machine with its client ID and client secret
  4. Look at the indexes on the universal auth settings table (identity_universal_auths) in the database the product's migrations built
s5-universal-auth-login-scans-client-ids
e2e/specs/identity-auth/s5-universal-auth-login-scans-client-ids.spec.ts:10
lines
infisical/backend/src/services/identity-ua/identity-ua-service.ts:116
infisical/backend/src/db/migrations/20231228074908_identity-universal-auth.ts:10
infisical/backend/src/db/migrations/20250710022434_add-index-for-access-token.ts:15
before the PRs testedno
Machine logins look through every identity's client secrets to check one
s6-client-secret-lookup-scans-every-identity-1
identity-authperformanceloss: time
medium

When A machine (a CI job, a service, an agent) logs in with Universal Auth, or an admin lists an identity's client secrets.

Got The product looks through the client secrets of every machine identity on the instance to find yours, so this gets slower as secrets are added anywhere.

Want Your identity's client secrets are found directly, at the same cost however many secrets exist.

steps
  1. In an organisation, create a machine identity and add Universal Auth to it
  2. Create a client secret for it
  3. Log in as the machine with its client ID and client secret
  4. Look at the indexes on the client secrets table (identity_ua_client_secrets) in the database the product's migrations built
s6-client-secret-lookup-scans-every-identity
e2e/specs/identity-auth/s6-client-secret-lookup-scans-every-identity.spec.ts:11
lines
infisical/backend/src/services/identity-ua/identity-ua-service.ts:162
infisical/backend/src/services/identity-ua/identity-ua-service.ts:951
infisical/backend/src/db/migrations/20231228074908_identity-universal-auth.ts:33
infisical/backend/src/db/migrations/20250710022434_add-index-for-access-token.ts:20
before the PRs testedno
An auth method that fails to load is shown as not set up on the identity
s5-auth-method-failed-load-shown-as-missing-1
identity-authbehaviourloss: decision
medium

When You open a machine identity's Universal Auth method and its settings fail to load.

Got The panel says no Universal Auth is associated with the identity, though it is set up.

Want The panel says the settings could not be loaded.

steps
  1. Create a machine identity and add Universal Auth to it
  2. Make the Universal Auth settings read fail with a server error (the client secrets list still loads)
  3. Open the identity's page and click Universal Auth in the Authentication card
  4. Wait past the one automatic retry
universal auth whose settings failed to load is not reported as missing
e2e/specs/identity-auth/s5-auth-method-failed-load-shown-as-missing.spec.ts:11
lines
infisical/frontend/src/views/IdentityAuthMethods/content/IdentityUniversalAuthContent.tsx:35
infisical/frontend/src/views/IdentityAuthMethods/content/IdentityUniversalAuthContent.tsx:47
infisical/frontend/src/views/IdentityAuthMethods/content/IdentityTokenAuthContent.tsx:24
infisical/frontend/src/views/IdentityAuthMethods/IdentityAuthMethodSheet.tsx:79
before the PRs testedno
Browser Back silently throws away unsaved Edit Sync changes
s10-edit-sync-draft-lost-on-browser-back-1
secret-syncsbehaviourloss: time
low

When You type changes in a sync's Edit Sync panel and press the browser's back button.

Got You land on the sync list with no warning, and your typed changes are gone when you return.

Want You are asked whether to discard your changes, as when you close the panel.

steps
  1. Open a project's Integrations page, Secret Syncs tab, then open an AWS Parameter Store sync's details page.
  2. Open Sync options > Edit Sync and type a description; do not save.
  3. Press the browser's back button.
  4. Press the browser's forward button and open Sync options > Edit Sync again.
lines
infisical/frontend/src/components/secret-syncs/EditSecretSyncModal.tsx:30
infisical/frontend/src/hooks/useDiscardChangesGuard.ts:11
infisical/frontend/src/pages/secret-manager/SecretSyncDetailsByIDPage/SecretSyncDetailsByIDPage.tsx:139
before the PRs testedno
A machine identity's page loads every client secret at once
s7-client-secret-list-unpaged-1
identity-authperformanceloss: time
medium

When A machine identity has many client secrets and you open its Universal Auth method.

Got The page loads every client secret at once though it shows five at a time, so it gets slower with each secret created.

Want The page loads only the page of client secrets you are looking at.

steps
  1. Sign in as an organization admin
  2. Create a machine identity and add Universal Auth to it
  3. Create 50 client secrets for it
  4. Open the identity's Universal Auth method and compare the client secrets the page loads with the five its Client Secrets table shows
s7-client-secret-list-unpaged
e2e/specs/identity-auth/s7-client-secret-list-unpaged.spec.ts:13
lines
infisical/backend/src/services/identity-ua/identity-ua-service.ts:951
infisical/backend/src/server/routes/v1/identity-universal-auth-router.ts:620
infisical/frontend/src/views/IdentityAuthMethods/helpers/IdentityUniversalAuthClientSecretsTable.tsx:130
infisical/backend/src/server/routes/v1/identity-token-auth-router.ts:446
before the PRs testedno
Opening a project's Service Tokens tab reads every project's service tokens on the instance
s5-service-token-list-reads-every-project-1
service-tokensperformanceloss: time
medium

When You open a project's Service Tokens tab.

Got The product looks through every project's service tokens on the instance to find your project's, so the tab slows as the instance grows.

Want It finds only your project's tokens directly.

steps
  1. Sign in as an admin of a Secrets Management project that has a few service tokens, on an instance where other projects have theirs
  2. Open the project's Access Management and pick the Service Tokens tab (it lists the project's tokens)
  3. Read the indexes of the service tokens table in Postgres
s5-service-token-list-reads-every-project
e2e/specs/service-tokens/s5-service-token-list-reads-every-project.spec.ts:10
lines
infisical/backend/src/db/migrations/20231225072545_service-token.ts:22
infisical/backend/src/services/service-token/service-token-service.ts:218
infisical/backend/src/server/routes/v1/project-router.ts:917
before the PRs testedno
The Service Tokens tab downloads every token of a project to show 20
s5-token-list-loads-every-token-1
service-tokensperformanceloss: time
medium

When Your project has many service tokens and you open the Service Tokens tab.

Got The page downloads every token in the project, then shows 20.

Want The page loads only the 20 tokens it shows.

steps
  1. Sign in as the admin of a project with more than 20 service tokens (25 in the repro)
  2. Open the project's Access Management on the Service Tokens tab; the table shows 20 per page
  3. Watch the token list the page downloads
s5-token-list-loads-every-token
e2e/specs/service-tokens/s5-token-list-loads-every-token.spec.ts:12
lines
infisical/backend/src/services/service-token/service-token-service.ts:218
infisical/backend/src/server/routes/v1/project-router.ts:921
infisical/frontend/src/hooks/api/serviceTokens/queries.tsx:17
infisical/frontend/src/pages/project/AccessControlPage/components/ServiceTokenTab/components/ServiceTokenSection/ServiceTokenTable.tsx:241
before the PRs testedno
A project's service token list looks empty when it fails to load
s1-service-token-list-failure-shown-as-empty-1
uiloss: decision
medium

When

Got

Want

steps
  1. Sign in as an admin of a project that has service tokens.
  2. Make the request that loads the project's service tokens fail on the server (the first try and its automatic retry).
  3. Open the project's Access Management, Service Tokens tab.
lines
infisical/frontend/src/pages/project/AccessControlPage/components/ServiceTokenTab/components/ServiceTokenSection/ServiceTokenTable.tsx:68
infisical/frontend/src/pages/project/AccessControlPage/components/ServiceTokenTab/components/ServiceTokenSection/ServiceTokenTable.tsx:170
infisical/frontend/src/hooks/api/serviceTokens/queries.tsx:26
before the PRs testedno
Revoking a service token at the same moment as another admin shows "Something went wrong"
s5-concurrent-token-revoke-server-error-1
service-tokensbehaviourloss: time
low

When You revoke a service token at the same moment as another admin does.

Got The token is revoked, but one of you is told "Something went wrong".

Want The second admin is told the token no longer exists.

steps
  1. Sign in as a project admin and create a service token in the project.
  2. Send two revokes of that token so both reach the server at the same moment (the test holds the token's row so both pass the existence check, then lets them go).
  3. Read both answers.
s5-concurrent-token-revoke-server-error
e2e/specs/service-tokens/s5-concurrent-token-revoke-server-error.spec.ts:8
lines
infisical/backend/src/services/service-token/service-token-service.ts:170
infisical/backend/src/services/service-token/service-token-service.ts:183
infisical/backend/src/lib/knex/index.ts:302
infisical/backend/src/server/routes/v2/service-token-router.ts:165
before the PRs testedno
A token deleted elsewhere stays listed, and deleting it fails every time
s5-deleted-elsewhere-token-stays-listed-1
service-tokensbehaviourloss: time
medium

When A token you see in the Service Tokens table was already deleted by someone else, and you delete it too.

Got You get an error and the token stays in the table. Every retry fails the same way until you reload.

Want The table refreshes and the deleted token disappears.

steps
  1. Sign in as a project admin and open the project's Access Management › Service Tokens tab with a token listed
  2. In another tab (or as another admin), delete that token
  3. Back in the first tab, without reloading, open the token's actions, choose Delete Token, type its name and confirm
  4. See the not-found error, then click Cancel
s5-deleted-elsewhere-token-stays-listed
e2e/specs/service-tokens/s5-deleted-elsewhere-token-stays-listed.spec.ts:8
lines
infisical/frontend/src/hooks/api/serviceTokens/queries.tsx:61
infisical/frontend/src/pages/project/AccessControlPage/components/ServiceTokenTab/components/ServiceTokenSection/ServiceTokenSection.tsx:70
infisical/backend/src/services/service-token/service-token-service.ts:171
before the PRs testedno
A service token cannot be scoped to an environment with a long slug
s6-token-scope-rejects-long-environment-slug-1
service-tokensbehaviourloss: time
medium

When You create a service token for an environment whose short name (slug) is longer than 50 characters.

Got The form marks the environment as too long and will not create the token, though it is a normal environment of the project.

Want The token is created for any environment the project has.

steps
  1. In a project's Settings, add an environment whose slug is 51 to 64 characters long (for example 60).
  2. Open Access Control and pick the Service Tokens tab.
  3. Click Create Token and type a name.
  4. Choose the long-slug environment under Environment.
  5. Click Create.
s6-token-scope-rejects-long-environment-slug
e2e/specs/service-tokens/s6-token-scope-rejects-long-environment-slug.spec.ts:8
lines
infisical/frontend/src/pages/project/AccessControlPage/components/ServiceTokenTab/components/ServiceTokenSection/AddServiceTokenModal.tsx:66
infisical/backend/src/server/routes/v1/project-env-router.ts:141
infisical/backend/src/server/routes/v2/service-token-router.ts:93
infisical/frontend/src/pages/secret-manager/SettingsPage/components/EnvironmentSection/AddEnvironmentModal.tsx:36
before the PRs testedno
The "Sync Failed" badge's list of failed syncs also shows syncs that succeeded
s11-sync-badge-popover-lists-every-sync-1
secret-syncsuiloss: nothing
low

When One of an environment's syncs fails and you open the "Sync Failed" badge on its secrets page.

Got The list titled "Failed Syncs" also shows syncs that succeeded or never ran.

Want The list shows only the syncs that failed.

steps
  1. Create a secrets project and add a secret to its dev environment.
  2. Create two AWS Parameter Store syncs from dev /: one to path /<name>/, one to path /ssm/<name>/ (a prefix AWS reserves), and trigger both.
  3. Wait until the first reads Synced and the second Failed to Sync.
  4. Open the dev environment's secrets page.
  5. Click the "Sync Failed" badge in the secrets table header.
s11-sync-badge-popover-lists-every-sync
e2e/specs/secret-syncs/s11-sync-badge-popover-lists-every-sync.spec.ts:53
lines
infisical/frontend/src/pages/secret-manager/OverviewPage/components/SecretSyncStatusBadgeOverview/SecretSyncStatusBadgeOverview.tsx:145
infisical/frontend/src/pages/secret-manager/OverviewPage/components/SecretSyncStatusBadgeOverview/SecretSyncStatusBadgeOverview.tsx:68
before the PRs testedno
A failed audit log stream list shows "No log streams have been configured" instead of an error
s3-log-stream-list-failure-shown-as-none-configured-1
behaviourloss: decision
medium

When

Got

Want

steps
  1. Sign in as an organization admin whose organization has at least one audit log stream.
  2. Make the log stream list request fail with a server error (500).
  3. Open Audit Logs > External Log Streams for the organization.
  4. Wait for the tab to settle.
lines
infisical/frontend/src/pages/organization/AuditLogsPage/components/AuditLogStreamTab/components/AuditLogStreamTable.tsx:56
infisical/frontend/src/pages/organization/AuditLogsPage/components/AuditLogStreamTab/components/AuditLogStreamTable.tsx:231
infisical/frontend/src/hooks/api/auditLogStreams/queries.tsx:52
before the PRs testedno
A machine token's use limit can be exceeded by renewing it mid-use
s9-renew-mid-use-gives-back-spent-use-1
identity-authbehaviourloss: work
medium

When You renew a machine token at the same moment it is used for another request.

Got The token works for one more request than its use limit allows.

Want The token stops working once its use limit is reached, even across a renew.

steps
  1. Create a machine identity with Universal Auth and an access-token use limit of 3, and log in to get token A (3 uses remaining).
  2. Call an identity-authenticated API once with A, leaving 2 uses remaining.
  3. Renew A to get token B for the same identity (uses remaining unchanged at 2).
  4. Start renewing B and, while its revocation read is in progress, call the API once with A so its use is counted (1 use remaining).
  5. Let the renew of B finish — it writes the remaining count it read earlier (2) back, overwriting the 1.
  6. Call the API with A repeatedly until it is refused, counting the successful calls.
s9 renew mid-use gives back a spent use: a 3-use token authorizes at most 3 requests
e2e/specs/identity-auth/s9-renew-mid-use-gives-back-spent-use.spec.ts:60
lines
infisical/backend/src/services/identity-access-token/identity-access-token-service.ts:591
infisical/backend/src/services/identity-access-token/identity-access-token-service.ts:640
infisical/backend/src/services/identity-access-token/identity-access-token-service.ts:641
infisical/backend/src/services/identity-access-token/identity-access-token-service.ts:502
before the PRs testedno
The audit logs' "3 months" range is refused as "Dates must be within 3 months" on a month's last day
s2-three-month-range-refused-at-month-end-1
behaviourloss: time
medium

When

Got

Want

steps
  1. Sign in as an organization admin on a month's last day whose date three months earlier falls in a shorter month (e.g. July 31 or December 31)
  2. Open Organization Audit Logs
  3. Open the range picker, pick the quick range "3 months" and press Apply
  4. A "Validation Error" toast flashes "dates must be within 3 months" and the panel shows "No audit logs on file"
lines
infisical/frontend/src/components/v3/platform/DateRangeFilter/DateRangeFilter.tsx:169
infisical/backend/src/server/routes/v1/organization-router.ts:284
before the PRs testedno
Renaming a custom log stream's header silently deletes the header
s6-renamed-stream-header-dropped-1
audit-logsbehaviourloss: work
high

When You rename a header on a custom log stream and leave its hidden value as is.

Got The save says it worked, but the header is gone, so the collector stops receiving it.

Want The renamed header keeps its value, or you are asked to type the value again.

steps
  1. Open Organization > Audit Logs > External Log Streams
  2. Add a Custom log stream with an endpoint URL and one header X-Heal with a value, and create it
  3. Open the stream's Edit form
  4. Change only the header name to X-Heal-Renamed, leaving the masked value untouched
  5. Click Update — the form closes as saved
  6. Open the stream's Edit form again
A custom log stream's renamed header keeps its value after the save
e2e/specs/audit-logs/s6-renamed-stream-header-dropped.spec.ts:6
lines
infisical/backend/src/ee/services/audit-log-stream/audit-log-stream-service.ts:154
infisical/frontend/src/pages/organization/AuditLogsPage/components/AuditLogStreamTab/AuditLogStreamForm/CustomProviderAuditLogStreamForm.tsx:111
before the PRs testedno
Add Log Stream offers only IBM QRadar when the provider list fails to load
s6-log-stream-provider-options-failure-shown-as-qradar-only-1
audit-logsbehaviourloss: time
medium

When You open Add Log Stream while the list of log providers cannot be loaded.

Got The picker shows only IBM QRadar, with no error, and choosing it asks you to upgrade.

Want The picker tells you the providers could not be loaded and lets you try again.

steps
  1. Sign in as an organization admin and open Audit Logs, External Log Streams tab.
  2. Make the log stream provider options request fail with a server error.
  3. Click Add Log Stream.
a provider options failure is not shown as a QRadar-only provider picker
e2e/specs/audit-logs/s6-log-stream-provider-options-failure-shown-as-qradar-only.spec.ts:6
lines
infisical/frontend/src/pages/organization/AuditLogsPage/components/AuditLogStreamTab/components/LogStreamProviderSelect.tsx:19
infisical/frontend/src/pages/organization/AuditLogsPage/components/AuditLogStreamTab/components/LogStreamProviderSelect.tsx:24
infisical/frontend/src/hooks/api/auditLogStreams/queries.tsx:30
before the PRs testedno
Filtering the audit logs by actor, event or secret reads every event in the range
s7-audit-log-filters-scan-every-event-1
audit-logsperformanceloss: time
medium

When You filter your organization's audit logs by a person, an event type or a secret.

Got The search goes through every event your organization recorded in the chosen range, so it slows as your history grows.

Want Filtered searches stay quick however many events the organization has.

steps
  1. Run Infisical with its own migrations and no ClickHouse (the default self-hosted install)
  2. Let an organization record many audit events (every secret read, write and audit-log view adds one)
  3. Open the organization's Audit Logs page and filter by an actor, an event type, or a secret path/key
  4. Read the audit_logs indexes, or EXPLAIN the list query the filter produces
audit log filter fields are each served by an index
e2e/specs/audit-logs/s7-audit-log-filters-scan-every-event.spec.ts:5
lines
infisical/backend/src/ee/services/audit-log/audit-log-dal.ts:167
infisical/backend/src/ee/services/audit-log/audit-log-dal.ts:174
infisical/backend/src/ee/services/audit-log/audit-log-dal.ts:185
infisical/backend/src/ee/services/audit-log/audit-log-dal.ts:210
infisical/backend/src/db/migrations/20240429154610_audit-log-index.ts:11
infisical/backend/src/db/manual-migrations/partition-audit-logs.ts:88
before the PRs testedno
An organisation's auth templates look missing when they fail to load
s7-auth-templates-load-failure-shown-as-empty-1
uiloss: decision
medium

When

Got

Want

steps
  1. Sign in as the admin of an organisation on the enterprise plan that holds a machine identity auth template (OIDC or LDAP).
  2. Open Organization Access Control and its Machine Identities tab while the server fails to answer the auth templates list request.
  3. Look at the Machine Identity Auth Templates card.
lines
infisical/frontend/src/pages/organization/AccessManagementPage/components/OrgIdentityTab/components/IdentitySection/IdentityAuthTemplatesTable.tsx:111
infisical/frontend/src/pages/organization/AccessManagementPage/components/OrgIdentityTab/components/IdentitySection/IdentityAuthTemplatesTable.tsx:139
infisical/frontend/src/pages/organization/AccessManagementPage/components/OrgIdentityTab/components/IdentitySection/IdentityAuthTemplatesTable.tsx:165
infisical/frontend/src/pages/organization/AccessManagementPage/components/OrgIdentityTab/components/IdentitySection/IdentitySection.tsx:255
before the PRs testedno
Adding a group to a project at the same moment twice fails with a server error
s13-group-added-to-project-twice-at-once-server-error-1
groupsbehaviourloss: time
low

When Two admins add the same group to a project at the same moment.

Got One of them is told "Something went wrong", though the group was added.

Want The second is told the group is already a member.

steps
  1. Sign in as an organisation admin and create a new group.
  2. Send two requests adding that group to the same project at the same moment (forced in the test by holding the group's row so both pass the membership check before either inserts) — through either the deprecated project-groups route or the project Access Control's memberships route the screen uses.
  3. Read both answers.
lines
infisical/infisical/backend/src/services/membership-group/membership-group-service.ts:176
infisical/infisical/backend/src/services/membership-group/membership-group-service.ts:190
infisical/infisical/backend/src/lib/knex/index.ts:219
before the PRs testedno
Cancelling Create Role throws away the typed role name without asking
s12-add-org-role-draft-discarded-on-cancel-1
roles-permissionsuiloss: time
low

When You start creating an organization role, type its name, then press Cancel.

Got The dialog closes straight away and the name you typed is gone when you open it again.

Want You are asked whether to discard what you typed, or the name is kept.

steps
  1. Sign in as an organization admin and open Access Control, Roles tab
  2. Click Create and type a role name
  3. Press Cancel
  4. Click Create again
s12-add-org-role-draft-discarded-on-cancel
e2e/specs/roles-permissions/s12-add-org-role-draft-discarded-on-cancel.spec.ts:8
lines
infisical/frontend/src/pages/organization/RoleByIDPage/components/RoleModal.tsx:145
infisical/frontend/src/pages/organization/RoleByIDPage/components/RoleModal.tsx:206
before the PRs testedno
Cancelling a new project role throws away the typed name without asking
s12-project-role-draft-discarded-on-cancel-1
roles-permissionsuiloss: time
low

When You start creating a project role, type its name, then press Cancel.

Got The sheet closes straight away and the name you typed is gone when you open it again.

Want You are asked whether to discard what you typed, or the name is kept.

steps
  1. Sign in as an organization admin and open a secret-management project's Access Control, Roles tab
  2. Press Add Project Role
  3. Type a role name
  4. Press Cancel
  5. Press Add Project Role again
s12-project-role-draft-discarded-on-cancel
e2e/specs/roles-permissions/s12-project-role-draft-discarded-on-cancel.spec.ts:10
lines
infisical/frontend/src/pages/project/RoleDetailsBySlugPage/components/RoleModal.tsx:108
infisical/frontend/src/pages/project/RoleDetailsBySlugPage/components/RoleModal.tsx:191
before the PRs testedno
A failed webhook list load shows a blank webhooks table with no error
s1-webhook-list-failure-shown-as-empty-table-1
behaviourloss: decision
medium

When

Got

Want

steps
  1. Sign in as an organization admin and open a project.
  2. Make the project's webhook list request fail with a server error.
  3. Open the project's Settings and select the Webhooks tab.
  4. Wait for the load and its single automatic retry to fail.
lines
infisical/frontend/src/pages/secret-manager/SettingsPage/components/WebhooksTab/WebhooksTab.tsx:86
infisical/frontend/src/pages/secret-manager/SettingsPage/components/WebhooksTab/WebhooksTab.tsx:210
infisical/frontend/src/hooks/api/webhooks/query.tsx:21
before the PRs testedno
A failed load of a project's Slack/Teams configuration shows "No workflow integrations configured"
s2-project-workflow-integration-failure-shown-as-unconfigured-1
behaviourloss: time
medium

When

Got

Want

steps
  1. Sign in as a project admin and create (or open) a secret-management project.
  2. Make the project's Slack and Microsoft Teams configuration reads fail with a server error.
  3. Open the project's Settings on the Workflow Integrations tab.
  4. Observe the tab once both reads (and their one retry) have failed.
lines
infisical/frontend/src/pages/secret-manager/SettingsPage/components/WorkflowIntegrationSection/WorkflowIntegrationTab.tsx:58
infisical/frontend/src/pages/secret-manager/SettingsPage/components/WorkflowIntegrationSection/WorkflowIntegrationTab.tsx:128
infisical/frontend/src/hooks/api/projects/queries.tsx:808
before the PRs testedno
Testing a webhook whose delivery fails still says "Successfully triggered webhook"
s1-webhook-test-failure-shown-as-success-1
behaviourloss: decision
medium

When

Got

Want

steps
  1. As a project admin, add a webhook on the project's Settings > Webhooks tab pointing at an address nothing listens on (e.g. http://127.0.0.1:1/hook)
  2. Open the row's actions menu and choose Test
  3. Watch the message that appears and the row's Status column
lines
infisical/backend/src/services/webhook/webhook-service.ts:207
infisical/backend/src/services/webhook/webhook-service.ts:213
infisical/backend/src/server/routes/v1/webhook-router.ts:281
infisical/frontend/src/pages/secret-manager/SettingsPage/components/WebhooksTab/WebhooksTab.tsx:178
before the PRs testedno
Finding a project's webhooks reads every webhook on the instance
s6-webhook-lookup-reads-every-webhook-1
webhooks-notificationsperformanceloss: time
medium

When You open a project's Webhooks tab, or change a secret in a project.

Got The product looks through every webhook on the instance to find your project's, so both slow as the instance grows.

Want It finds only your project's webhooks directly.

steps
  1. Sign in as an admin of a Secrets Management project, on an instance where other projects have webhooks
  2. Create one webhook on the project's Settings, Webhooks tab
  3. Reopen the Webhooks tab (or change a secret in the project, which looks up its webhooks to trigger them)
  4. Read the indexes of the webhooks table in Postgres
s6-webhook-lookup-reads-every-webhook
e2e/specs/webhooks-notifications/s6-webhook-lookup-reads-every-webhook.spec.ts:8
lines
infisical/backend/src/db/migrations/20231225072552_webhook.ts:22
infisical/backend/src/services/webhook/webhook-dal.ts:67
infisical/backend/src/services/webhook/webhook-fns.ts:710
infisical/backend/src/services/webhook/webhook-service.ts:231
before the PRs testedno
Project Slack setup says no Slack workspace is connected when the list failed to load
s6-project-slack-form-org-list-failure-shown-as-none-connected-1
webhooks-notificationsbehaviourloss: time
medium

When You set up Slack notifications for a project while the organization's integrations fail to load.

Got The form says no Slack workspace is connected and sends you to connect one, with no error.

Want The form says the integrations could not be loaded.

steps
  1. Sign in as an org admin and open a project's Settings on the Workflow Integrations tab.
  2. Make the organization's workflow integration list request fail with a server error.
  3. Click Add integration and pick Slack.
  4. Read the Slack form.
s6-project-slack-form-org-list-failure-shown-as-none-connected
e2e/specs/webhooks-notifications/s6-project-slack-form-org-list-failure-shown-as-none-connected.spec.ts:6
lines
infisical/frontend/src/hooks/api/workflowIntegrations/queries.tsx:151
infisical/frontend/src/pages/secret-manager/SettingsPage/components/WorkflowIntegrationSection/components/SlackIntegrationForm.tsx:111
infisical/frontend/src/pages/secret-manager/SettingsPage/components/WorkflowIntegrationSection/components/SlackIntegrationForm.tsx:222
before the PRs testedno
Project machine identity page fails two accessibility checks
s13-project-identity-page-a11y-violations-1
machine-identitiesuiloss: nothing
low

When You use a screen reader, keyboard or need strong contrast on a machine identity's page inside a project.

Got Each role row announces itself as a button while holding another button, and the grey section descriptions are slightly too faint.

Want Role rows and their actions are separate controls, and the descriptions are easy to read.

steps
  1. Sign in as an organisation admin.
  2. Add a machine identity to a secret-management project with a project role (for example Viewer).
  3. Open the machine identity's page inside the project and wait until the role table and the section descriptions are shown.
  4. Run an accessibility check (axe) on the page's content, outside the sidebar and header.
s13-project-identity-page-a11y-violations
e2e/specs/machine-identities/s13-project-identity-page-a11y-violations.spec.ts:11
lines
infisical/frontend/src/pages/project/IdentityDetailsByIDPage/components/IdentityRoleDetailsSection/IdentityRoleDetailsSection.tsx:245
infisical/frontend/src/pages/project/IdentityDetailsByIDPage/components/IdentityRoleDetailsSection/IdentityRoleDetailsSection.tsx:304
infisical/frontend/src/components/v3/generic/Card/Card.tsx:49
before the PRs testedno
A server failure on the sign-up code is shown as a wrong code and clears it
s2-signup-code-server-error-blames-code-1
behaviourloss: time
medium

When

Got

Want

steps
  1. Open the sign-up page signed out and enter a new email, then continue with email
  2. Open the verification email and read the six-digit code
  3. Have the sign-up code check fail on the server (a server error answer) or lose the connection as it is sent
  4. Type the correct code into the six code boxes
lines
infisical/frontend/src/components/auth/CodeInputStep.tsx:69
infisical/frontend/src/components/auth/CodeInputStep.tsx:88
infisical/frontend/src/components/v3/platform/VerificationCode/VerificationCodeForm.tsx:85
before the PRs testedno
A failed notification load leaves the notifications panel blank, with no error and no unread mark on the bell
s1-notification-load-failure-shown-blank-1
behaviourloss: decision
medium

When

Got

Want

steps
  1. Sign in and land on any organization page.
  2. Make the user's notifications request fail with a server error.
  3. Wait for the load and its single automatic retry to fail.
  4. Click the Notifications bell in the top navbar.
lines
infisical/frontend/src/layouts/OrganizationLayout/components/NavBar/NotificationDropdown.tsx:41
infisical/frontend/src/layouts/OrganizationLayout/components/NavBar/NotificationDropdown.tsx:135
infisical/frontend/src/layouts/OrganizationLayout/components/NavBar/NotificationDropdown.tsx:144
infisical/frontend/src/hooks/api/notifications/queries.tsx:13
before the PRs testedno
Editing a log stream's URL can undo another admin's product filter
s8-stale-stream-edit-wipes-product-filter-1
audit-logsbehaviourloss: work
low

When You change a log stream's URL in an edit form you opened before another admin limited its products.

Got Your save quietly removes their product filter, and the stream sends every product's events again.

Want Only the URL changes, or you are told the stream changed since you opened the form.

steps
  1. As an org admin, create a Custom audit log stream with Products left at All products.
  2. Open the stream's Edit form in a second browser tab and leave it open.
  3. In the first tab, open Edit on the same stream, set Products to Secret Management and click Update.
  4. In the second tab, change only the Endpoint URL and click Update.
  5. Open the stream's Edit form again.
A log stream's product filter survives a URL save from an edit form opened earlier
e2e/specs/audit-logs/s8-stale-stream-edit-wipes-product-filter.spec.ts:27
lines
infisical/frontend/src/pages/organization/AuditLogsPage/components/AuditLogStreamTab/AuditLogStreamForm/AuditLogStreamForm.tsx:32
infisical/frontend/src/pages/organization/AuditLogsPage/components/AuditLogStreamTab/AuditLogStreamForm/AuditLogStreamForm.tsx:82
infisical/backend/src/ee/services/audit-log-stream/audit-log-stream-service.ts:100
infisical/backend/src/ee/services/audit-log-stream/audit-log-stream-service.ts:205
#8526
reviewed by greptile, cubic, coderabbit
no
A secret request link that fails to load tells the requested person the request is missing or has expired
s3-secret-request-load-failure-shown-as-expired-1
behaviourloss: decision
medium

When

Got

Want

steps
  1. Sign in as a member and open Secret Sharing, Request Secret tab
  2. Create a secret request and copy its link
  3. Open the link while loading the request fails with a server error
  4. Read the alert on the page
lines
infisical/frontend/src/pages/public/ViewSecretRequestByIDPage/ViewSecretRequestByIDPage.tsx:193
infisical/frontend/src/pages/public/ViewSecretRequestByIDPage/components/SecretErrorContainer.tsx:27
infisical/frontend/src/pages/public/ViewSecretRequestByIDPage/ViewSecretRequestByIDPage.tsx:57
before the PRs testedno
Sign-up starts over after the browser's back and forward buttons
s13-signup-restarts-after-back-forward-1
project-access-controlbehaviourloss: time
low

When You have verified your email and filled in the account form, then press back and forward in the browser.

Got Nothing warns you, and sign-up starts over at the email step with your details gone.

Want You are warned before leaving, or you come back to the filled-in account form.

steps
  1. Signed out, open the login page, then go to the sign-up page.
  2. Enter a fresh email and request a code.
  3. Enter the code from the email.
  4. Fill in the account form: name, organisation name, password and its confirmation; do not submit.
  5. Press the browser's back button and wait for the login page.
  6. Press the browser's forward button.
the verified sign-up survives browser back then forward from the account form
e2e/specs/project-access-control/s13-signup-restarts-after-back-forward.spec.ts:11
lines
infisical/frontend/src/pages/auth/SignUpPage/SignUpPage.tsx:60
before the PRs testedno
A secret request accepts two answers when they arrive together
s8-secret-request-two-answers-both-accepted-1
secret-sharingbehaviourloss: work
medium

When Two people answer the same secret request at the same moment.

Got Both are told their secret was sent. The requester gets two completion emails, but only one of the two values is kept.

Want One answer is kept and the other person is told the request was already answered.

steps
  1. Sign in and create a secret request that anyone with the link may answer.
  2. Answer it once, then answer it again: the second answer is refused with 'Secret request already has a value set'.
  3. Create a fresh request the same way.
  4. Have two people send different values to it at the same moment (the test holds the request's row so both reach the write).
s8-secret-request-two-answers-both-accepted
e2e/specs/secret-sharing/s8-secret-request-two-answers-both-accepted.spec.ts:9
lines
infisical/backend/src/services/secret-sharing/secret-sharing-service.ts:445
infisical/backend/src/services/secret-sharing/secret-sharing-service.ts:481
infisical/backend/src/services/secret-sharing/secret-sharing-service.ts:499
infisical/backend/src/services/secret-sharing/secret-sharing-dal.ts:16
before the PRs testedno
Shared-secret and secret-request lists slow down as every organisation's shares grow
s8-shared-secrets-list-scans-every-share-1
secret-sharingperformanceloss: time
medium

When You open your shared secrets or your secret requests.

Got The product looks through every shared secret of every organisation to find yours, so the list gets slower as the instance is used.

Want It finds only your own shares directly, however many others exist.

steps
  1. Sign in as a member of an organisation
  2. Share a secret from Secret Sharing
  3. Open the Share Secrets list (or the Request Secret list)
  4. Read the index catalog of the secret_sharing table in the database
s8-shared-secrets-list-scans-every-share
e2e/specs/secret-sharing/s8-shared-secrets-list-scans-every-share.spec.ts:10
lines
infisical/backend/src/db/migrations/20240528190137_secret_sharing.ts:8
infisical/backend/src/db/migrations/20240925100349_managed-secret-sharing.ts:23
infisical/backend/src/services/secret-sharing/secret-sharing-service.ts:574
infisical/backend/src/services/secret-sharing/secret-sharing-dal.ts:65
before the PRs testedno
Opening a publicly shared secret reports the secret as missing while loading its branding
s8-public-share-branding-reports-secret-missing-1
secret-sharingbehaviourloss: time
low

When You open a link to a secret someone shared from the public share page.

Got The page asks for the secret's branding and is told the secret does not exist, then asks again. The secret still opens, after the extra wait.

Want The page gets the default look straight away for a secret that exists.

steps
  1. Without signing in, open the public share page (/shared/new).
  2. Type a secret value and create the secret link.
  3. Open the link and reveal the secret.
  4. Watch the branding lookup the page makes for that secret.
s8-public-share-branding-reports-secret-missing
e2e/specs/secret-sharing/s8-public-share-branding-reports-secret-missing.spec.ts:9
lines
infisical/backend/src/server/routes/v1/secret-sharing-router.ts:596
infisical/backend/src/services/secret-sharing/secret-sharing-service.ts:847
infisical/backend/src/services/secret-sharing/secret-sharing-service.ts:537
infisical/frontend/src/pages/public/ViewSharedSecretByIDPage/ViewSharedSecretByIDPage.tsx:209
before the PRs testedno
An empty password on a share link sends the recipient to the sign-in page
s8-empty-password-share-link-sends-to-login-1
secret-sharingbehaviourloss: time
medium

When You open a password-protected secret link that anyone may open, and press Enter before typing the password.

Got Infisical tells you a login is required and takes you to its sign-in page, as if you needed an account.

Want The link stays open and asks you for the password.

steps
  1. Signed in, open Secret Sharing, click Share Secret, enter a value, open More Options, set a password and turn off 'Limit access to people within organization', create the link
  2. In a browser that is not signed in, open the link
  3. With the password field empty, press Enter
an empty password on a protected link shared with anyone keeps the recipient on the link
e2e/specs/secret-sharing/s8-empty-password-share-link-sends-to-login.spec.ts:9
lines
infisical/frontend/src/pages/public/ViewSharedSecretByIDPage/components/PasswordContainer.tsx:18
infisical/frontend/src/pages/public/ViewSharedSecretByIDPage/components/PasswordContainer.tsx:62
infisical/frontend/src/hooks/api/secretSharing/mutations.ts:41
infisical/backend/src/services/secret-sharing/secret-sharing-service.ts:751
infisical/frontend/src/pages/public/ViewSharedSecretByIDPage/ViewSharedSecretByIDPage.tsx:108
before the PRs testedno
Deleting a shared secret logs a loading-animation error in the console
s5-shared-secret-delete-logs-loader-error-1
uiloss: nothing
low

When

Got

Want

steps
  1. Sign in and open Secret Sharing, Share Secret tab
  2. Share a secret with a name so it appears in the Shared Secrets list
  3. Open the row's actions, choose Delete, and confirm Delete in the dialog
  4. Watch the browser console
lines
infisical/frontend/src/pages/organization/SecretSharingPage/components/ShareSecret/ShareSecretTab.tsx:93
infisical/frontend/src/components/v3/generic/Button/Button.tsx:111
infisical/frontend/src/components/v3/generic/Loader/Loader.tsx:96
before the PRs testedno
Product descriptions on the Projects page are too faint to read easily
s10-projects-page-product-descriptions-low-contrast-1
identity-authuiloss: nothing
low

When You open your organisation's Projects page.

Got The short description under each product's title is a faint grey on the dark card, slightly below the readable minimum.

Want The descriptions are dark-on-light enough to read comfortably.

steps
  1. Sign in as an organisation admin
  2. Open the app's home address; it sends you to the organisation's Projects page
  3. Look at the short description under each product card's title (for example 'Centralize secrets across environments…')
  4. Check its colour contrast against the card background
the product descriptions on the Projects page meet the contrast minimum
e2e/specs/identity-auth/s10-projects-page-product-descriptions-low-contrast.spec.ts:6
lines
infisical/frontend/src/pages/organization/ProjectsPage/components/ProjectCategoryOverview.tsx:560
before the PRs testedno
Past the 100th shared secret, the next page of the Shared Secrets list fails to load and the pager disappears
s1-shared-secrets-past-100-cannot-be-listed-1
behaviourloss: time
medium

When

Got

Want

steps
  1. Sign in as a member with more than 110 shared secrets in their organisation (create them with Share Secret)
  2. Open Organization > Secret Sharing, Share Secret tab, at 10 rows per page
  3. Click Go to last page (page 12, offset 110)
  4. The table is replaced by 'Could Not Load Shared Secrets — Refresh the page to try again' and the pager is gone
lines
infisical/backend/src/server/routes/v1/secret-sharing-router.ts:51
infisical/backend/src/server/routes/v1/secret-requests-router.ts:252
infisical/frontend/src/pages/organization/SecretSharingPage/components/ShareSecret/ShareSecretsTable.tsx:38
infisical/frontend/src/pages/organization/SecretSharingPage/components/ShareSecret/ShareSecretsTable.tsx:79
infisical/frontend/src/pages/organization/SecretSharingPage/components/RequestSecret/RequestedSecretsTable.tsx:32
before the PRs testedno
Sharing a secret with a few emails loads every member of the organisation
s9-share-to-emails-reads-every-org-member-1
secret-sharingperformanceloss: time
low

When You share a secret with a few people's emails in an organisation with many members.

Got Infisical loads every member of the organisation to check those few addresses. The work grows with every member added.

Want Sharing looks up only the emails you chose.

steps
  1. Sign in to an organisation as a member.
  2. Create a shared secret restricted to the organisation, authorised for one member's email, counting the rows of the organisation-member query.
  3. Invite 45 more people to the organisation.
  4. Create another shared secret authorised for the same one email, counting the same query.
s9-share-to-emails-reads-every-org-member
e2e/specs/secret-sharing/s9-share-to-emails-reads-every-org-member.spec.ts:16
lines
infisical/backend/src/services/secret-sharing/secret-sharing-service.ts:201
infisical/backend/src/services/org/org-dal.ts:512
before the PRs testedno
Cancelling a new client secret throws the typed description away without a warning
s11-client-secret-description-wiped-on-cancel-1
identity-authuiloss: work
low

When You type a description for a new client secret, then cancel or close the dialog.

Got It closes with no warning, and the description is gone when you open it again.

Want Warn before discarding the description, or keep it.

steps
  1. Open a machine identity that has Universal Auth, in Access Control > Identities
  2. Open its Universal Auth method and choose Add Client Secret
  3. Type a description
  4. Choose Cancel (or the close button, Escape, or a click outside the dialog)
  5. Choose Add Client Secret again
s11-client-secret-description-wiped-on-cancel
e2e/specs/identity-auth/s11-client-secret-description-wiped-on-cancel.spec.ts:9
lines
infisical/infisical/frontend/src/pages/organization/IdentityDetailsByIDPage/components/IdentityClientSecretModal.tsx:74
infisical/infisical/frontend/src/pages/organization/IdentityDetailsByIDPage/components/IdentityClientSecretModal.tsx:164
before the PRs testedno
Cancelling the Create Token sheet throws away the typed service token name without asking
s3-create-token-cancel-discards-name-1
uiloss: time
low

When

Got

Want

steps
  1. Sign in as a project admin and open a secret-management project's Access Management, Service Tokens tab
  2. Press Create Token
  3. Type a name in Service Token Name
  4. Press Cancel
  5. Press Create Token again
lines
infisical/frontend/src/pages/project/AccessControlPage/components/ServiceTokenTab/components/ServiceTokenSection/AddServiceTokenModal.tsx:426
infisical/frontend/src/pages/project/AccessControlPage/components/ServiceTokenTab/components/ServiceTokenSection/AddServiceTokenModal.tsx:446
infisical/frontend/src/pages/project/AccessControlPage/components/ServiceTokenTab/components/ServiceTokenSection/AddServiceTokenModal.tsx:456
before the PRs testedno
Cancelling Create Machine Identity throws away the typed name without asking
s14-create-identity-draft-discarded-on-cancel-1
machine-identitiesuiloss: time
low

When You start creating a machine identity, type its name, then press Cancel.

Got The dialog closes straight away and the name you typed is gone when you open it again.

Want You are asked whether to discard what you typed, or the name is kept.

steps
  1. Sign in as an organization admin and open Access Control, Identities tab
  2. Click Create and type a machine identity name
  3. Press Cancel
  4. Click Create again
s14-create-identity-draft-discarded-on-cancel
e2e/specs/machine-identities/s14-create-identity-draft-discarded-on-cancel.spec.ts:8
lines
infisical/frontend/src/pages/organization/AccessManagementPage/components/OrgIdentityTab/components/IdentitySection/CreateOrgIdentitySheet.tsx:179
infisical/frontend/src/pages/organization/AccessManagementPage/components/OrgIdentityTab/components/IdentitySection/CreateOrgIdentitySheet.tsx:200
#8675
reviewed by greptile, cubic, coderabbit
no
Cancelling the edit dialog of a machine identity throws away the new name without asking
s15-identity-rename-discarded-on-cancel-1
machine-identitiesuiloss: time
low

When You change a machine identity's name in its edit dialog and press Cancel.

Got The dialog closes without asking. When you open it again, your new name is gone.

Want You are asked before the change is thrown away, or it is still there.

steps
  1. Sign in as an organization admin.
  2. Open a machine identity's detail page.
  3. Open the edit details dialog (Update Machine Identity).
  4. Change the name, do not save.
  5. Click Cancel.
  6. Open the edit details dialog again.
cancelling an edited machine identity name warns first or keeps the edit
e2e/specs/machine-identities/s15-identity-rename-discarded-on-cancel.spec.ts:7
lines
infisical/frontend/src/pages/organization/AccessManagementPage/components/OrgIdentityTab/components/IdentitySection/OrgIdentityModal.tsx:293
infisical/frontend/src/pages/organization/IdentityDetailsByIDPage/IdentityDetailsByIDPage.tsx:222
infisical/frontend/src/pages/organization/AccessManagementPage/components/OrgIdentityTab/components/IdentitySection/OrgIdentityModal.tsx:100
#8624
reviewed by greptile, cubic, coderabbit
no
Cancelling Add Machine Identity to Project drops the picked identity and role
s14-project-identity-picks-discarded-on-cancel-1
machine-identitiesuiloss: time
low

When You start adding an existing machine identity to a project, pick it and a role, then press Cancel.

Got The sheet closes straight away and your picks are gone when you open it again.

Want You are asked whether to discard your picks, or they are kept.

steps
  1. Sign in as an organization admin and open a secret-management project's Access Control, Machine Identities tab
  2. Press Add Machine Identity to Project and choose Assign Existing
  3. Pick an organization machine identity and the Viewer role
  4. Press Cancel
  5. Press Add Machine Identity to Project again and choose Assign Existing
s14-project-identity-picks-discarded-on-cancel
e2e/specs/machine-identities/s14-project-identity-picks-discarded-on-cancel.spec.ts:8
lines
infisical/frontend/src/pages/project/AccessControlPage/components/IdentityTab/components/CreateProjectIdentity/CreateProjectIdentityForm.tsx:408
infisical/frontend/src/pages/project/AccessControlPage/components/IdentityTab/components/CreateProjectIdentity/CreateProjectIdentitySheet.tsx:48
before the PRs testedno
Cancelling a new access token silently throws away the typed name
s10-access-token-name-wiped-on-cancel-1
identity-authbehaviourloss: time
low

When You type a name for a new access token and press Cancel.

Got The dialog closes with no warning, and when you open it again the name is gone.

Want You are warned first, or your typed name is still there.

steps
  1. Sign in as an organization admin.
  2. Open a machine identity that has Token Auth, and open its Token Auth method.
  3. Click Add Token and type a name into the Name field.
  4. Click Cancel.
  5. Click Add Token again.
s10-access-token-name-wiped-on-cancel
e2e/specs/identity-auth/s10-access-token-name-wiped-on-cancel.spec.ts:10
lines
infisical/frontend/src/pages/organization/IdentityDetailsByIDPage/components/IdentityTokenModal.tsx:145
infisical/frontend/src/pages/organization/IdentityDetailsByIDPage/components/IdentityTokenModal.tsx:80
before the PRs testedno
Cancelling the Create Group dialog keeps the typed group name, while Esc, the X or a click outside clear it
s7-create-group-cancel-keeps-draft-1
behaviourloss: nothing
low

When

Got

Want

steps
  1. As an org admin, open Access Management and the Groups tab
  2. Click Create Group and type a name and slug
  3. Click Cancel
  4. Click Create Group again: the cancelled name and slug are still filled in
  5. Type a name again and close with Esc (or the X, or a click outside), then reopen: the fields are empty
lines
infisical/frontend/src/pages/organization/AccessManagementPage/components/OrgGroupsTab/components/OrgGroupsSection/OrgGroupModal.tsx:198
infisical/frontend/src/pages/organization/AccessManagementPage/components/OrgGroupsTab/components/OrgGroupsSection/OrgGroupModal.tsx:216
infisical/frontend/src/pages/organization/AccessManagementPage/components/OrgGroupsTab/components/OrgGroupsSection/OrgGroupModal.tsx:103
before the PRs testedno
Cancelling Add Group Members drops the picked member with no warning
s15-group-member-pick-discarded-on-cancel-1
groupsuiloss: time
low

When You pick a member in Add Group Members, then press Cancel.

Got The dialog closes straight away and your pick is gone when you open it again.

Want You are asked whether to discard the pick, or it is kept.

steps
  1. Sign in as an organisation admin and open Access Management, Groups, then an empty group's detail page
  2. Press Add Member
  3. Pick a user in Members, so the button reads Assign 1 Member
  4. Press Cancel
  5. Press Add Member again
s15-group-member-pick-discarded-on-cancel
e2e/specs/groups/s15-group-member-pick-discarded-on-cancel.spec.ts:10
lines
infisical/frontend/src/pages/organization/GroupDetailsByIDPage/components/AddGroupMemberModal.tsx:287
infisical/frontend/src/pages/organization/GroupDetailsByIDPage/components/AddGroupMemberModal.tsx:179
before the PRs testedno
Listing a KMS project's keys reads every key on the instance
s12-project-key-list-reads-every-key-1
kmsperformanceloss: time
medium

When You open a KMS project's Keys table.

Got The product looks through every key on the whole instance to find your project's, so the list slows down as the instance grows.

Want It finds only your project's keys directly.

steps
  1. Sign in as an organisation admin on an instance where other organisations and projects hold their own KMS keys
  2. Create a KMS project and add two keys to it
  3. Open the project's Overview so the Keys table lists the project's keys
  4. Read the indexes of the kms_keys table in Postgres, and EXPLAIN the list statement
s12-project-key-list-reads-every-key
e2e/specs/kms/s12-project-key-list-reads-every-key.spec.ts:12
lines
infisical/backend/src/services/kms/kms-key-dal.ts:193
infisical/backend/src/services/kms/kms-key-dal.ts:156
infisical/backend/src/db/migrations/20241003220151_kms-key-cmek-alterations.ts:22
infisical/backend/src/db/migrations/20240603075514_kms.ts:17
before the PRs testedno
Every KMIP operation reads all certificates of every organisation to check its own
s12-kmip-certificate-lookup-unindexed-1
kmsperformanceloss: time
medium

When A KMIP client sends any operation to Infisical.

Got Before running it, Infisical reads through every KMIP certificate issued on the whole instance to find the two it needs, so each operation gets slower as more certificates are issued.

Want Infisical finds the two certificates directly, whatever the number of certificates.

steps
  1. Read the indexes of kmip_client_certificates and kmip_org_server_certificates in the sandbox database (the schema is built by the product's own migrations at boot).
  2. Note that each table has only its primary key on id.
  3. EXPLAIN the client-certificate lookup the KMIP session check makes: select * from kmip_client_certificates where "serialNumber" in (…) limit 1.
KMIP client certificates are found by serial number through an index
e2e/specs/kms/s12-kmip-certificate-lookup-unindexed.spec.ts:5
lines
infisical/backend/src/ee/services/kmip/kmip-service.ts:915
infisical/backend/src/ee/services/kmip/kmip-service.ts:923
infisical/backend/src/ee/routes/v1/kmip-spec-router.ts:61
infisical/backend/src/db/migrations/20250203141127_add-kmip.ts:76
before the PRs testedno
Exporting a key shows an empty dialog when its key material fails to load
s13-key-export-load-failure-shown-blank-1
kmsbehaviourloss: time
medium

When You open Export Key on a key and its key material cannot be loaded.

Got The dialog opens with just its title and nothing else, with no message.

Want The dialog tells you the key material could not be loaded.

steps
  1. Sign in as an org admin and open a KMS project's Project Overview holding an exportable RSA_4096 sign-verify key
  2. Have the key's private-key and public-key loads fail on the server
  3. Open the key's row menu and choose Export Key
  4. Wait for the loading spinner to go
a failed key-material load shows an error in the Export Key dialog
e2e/specs/kms/s13-key-export-load-failure-shown-blank.spec.ts:9
lines
infisical/frontend/src/pages/kms/OverviewPage/components/CmekExportKeyModal.tsx:51
infisical/frontend/src/pages/kms/OverviewPage/components/CmekExportKeyModal.tsx:80
infisical/frontend/src/pages/kms/OverviewPage/components/CmekExportKeyModal.tsx:94
infisical/frontend/src/hooks/api/cmeks/queries.tsx:62
before the PRs testedno
A deleted native integration's detail link says "Unable to fetch integration" instead of saying it no longer exists
s4-deleted-integration-detail-reads-as-fetch-error-1
uiloss: time
low

When

Got

Want

steps
  1. Open a Secrets Management project that has a native AWS Parameter Store integration and note its detail link
  2. Delete the integration (from the Native Integrations tab, another tab, or the API)
  3. Open the integration's detail link again
  4. Observe the page shows only "Error: Unable to fetch integration." — the same text it shows for a server error — and never says the integration was not found
lines
infisical/frontend/src/pages/secret-manager/IntegrationsDetailsByIDPage/IntegrationsDetailsByIDPage.tsx:42
infisical/frontend/src/pages/secret-manager/IntegrationsDetailsByIDPage/IntegrationsDetailsByIDPage.tsx:147
before the PRs testedno
Two edits of an external KMS saved at once lose one of the changes
s13-external-kms-concurrent-edits-lose-change-1
kmsbehaviourloss: work
medium

When Two people or scripts change different settings of the same external KMS at the same moment.

Got Both are told the change was saved, but one change is gone and the old value is back.

Want Both changes are kept, or one is told to retry.

steps
  1. Create two AWS external KMS entries through the API (the sandbox's LocalStack AWS KMS); note the key id the first one generated (K2).
  2. On a third AWS external KMS, send an edit with only provider.inputs.kmsKeyId = K2 and, at the same time, an edit with only provider.inputs.credential (access key test2). In the test both are held at their save by a row lock on the external_kms row and released together.
  3. Both edits answer 200.
  4. Read the third external KMS back by id.
s13-external-kms-concurrent-edits-lose-change: two edits of different settings both survive when their saves meet
e2e/specs/kms/s13-external-kms-concurrent-edits-lose-change.spec.ts:28
lines
infisical/backend/src/ee/services/external-kms/external-kms-service.ts:207
infisical/backend/src/ee/services/external-kms/external-kms-service.ts:228
infisical/backend/src/ee/services/external-kms/external-kms-service.ts:231
infisical/backend/src/ee/services/external-kms/external-kms-service.ts:284
infisical/backend/src/ee/services/external-kms/external-kms-service.ts:297
infisical/backend/src/ee/services/external-kms/providers/model.ts:118
before the PRs testedno
ECC P-384 and P-521 signing keys can't export their public key on the keys page
s13-ecc-p384-signing-key-export-disabled-1
kmsuiloss: time
medium

When You have an ECC P-384 or P-521 signing key that was created as non-exportable.

Got Its Export Key action is greyed out as non-exportable, so you can't get its public key from the keys page, while a P-256 key created the same way can.

Want Export Key shows the public key of every signing key, P-384 and P-521 included.

steps
  1. Through the API, create a KMS project and two sign-verify keys with isExportable false: one ECC_NIST_P256, one ECC_NIST_P384
  2. Open the project's Overview
  3. Open the P-256 key's row menu: Export Key is enabled
  4. Open the P-384 key's row menu: Export Key is disabled
a non-exportable ECC P-384 signing key's Export Key action is available for its public key
e2e/specs/kms/s13-ecc-p384-signing-key-export-disabled.spec.ts:9
lines
infisical/frontend/src/hooks/api/cmeks/types.ts:155
infisical/backend/src/lib/crypto/sign/types.ts:17
infisical/frontend/src/pages/kms/OverviewPage/components/CmekTable.tsx:514
infisical/frontend/src/pages/kms/OverviewPage/components/CmekExportKeyModal.tsx:35
infisical/frontend/src/pages/kms/OverviewPage/components/CmekBulkImportModal.tsx:156
before the PRs testedno
Audit logs that fail to load are shown as "No audit logs on file"
s1-org-audit-logs-load-failure-shown-as-empty-1
behaviourloss: decision
medium

When

Got

Want

steps
  1. Sign in as an organization admin.
  2. Make the audit logs read fail with a server error.
  3. Open the organization's Audit Logs page, or a native integration's detail page (its Integration Logs section).
  4. Wait for the error toast to fade: the panel says "No audit logs on file", with no error and no way to retry.
lines
infisical/frontend/src/hooks/api/auditLogs/queries.tsx:49
infisical/frontend/src/hooks/api/auditLogs/queries.tsx:53
infisical/frontend/src/pages/organization/AuditLogsPage/components/LogsTable.tsx:56
infisical/frontend/src/pages/secret-manager/IntegrationsDetailsByIDPage/components/IntegrationAuditLogsSection.tsx:50
before the PRs testedno
KMIP Locate sends back every key of a project in one answer
s14-kmip-locate-returns-every-key-1
kmsperformanceloss: time
medium

When A KMIP client looks up the keys of a project that holds many keys.

Got It gets every key of the project in a single answer, however many there are.

Want It gets the keys one page at a time, like the project's key list.

steps
  1. Create a KMS project and add 5 keys to it
  2. Create a machine identity with the org admin role, give it a token, and register an organisation KMIP server certificate with it
  3. Create a KMIP client in the project with the Locate permission and issue it a client certificate
  4. Send a KMIP Locate for the project with the client id, project id and both certificate serial numbers: 5 keys come back
  5. Add 150 more keys to the project (the key list now reports 155)
  6. Send the same KMIP Locate again
KMIP Locate answers at most one bounded page of a project's keys
e2e/specs/kms/s14-kmip-locate-returns-every-key.spec.ts:14
lines
infisical/backend/src/ee/services/kmip/kmip-operation-service.ts:407
infisical/backend/src/services/kms/kms-key-dal.ts:154
infisical/backend/src/ee/routes/v1/kmip-spec-router.ts:380
before the PRs testedno
The Certificate Authorities page says a project has no internal CAs when its CA list fails to load
s1-internal-ca-list-failure-shown-as-empty-1
behaviourloss: decision
medium

When

Got

Want

steps
  1. Sign in and open a Certificate Management project that holds an active internal root CA.
  2. Make the internal CA list request fail with a server error (GET /api/v1/cert-manager/ca/internal answers 500).
  3. Open the project's Certificate Authorities page on the Internal tab.
  4. Wait for the read to give up (it is retried once).
lines
infisical/frontend/src/pages/cert-manager/CertificateAuthoritiesPage/components/CaTable.tsx:57
infisical/frontend/src/pages/cert-manager/CertificateAuthoritiesPage/components/CaTable.tsx:70
infisical/frontend/src/hooks/api/ca/queries.tsx:56
before the PRs testedno
A failed load of external certificate authorities shows "No existing external certificate authorities" instead of an error
s1-external-ca-list-failure-shown-as-empty-1
behaviourloss: decision
medium

When

Got

Want

steps
  1. Sign in as a user who administers a Certificate Management project.
  2. Open the project's Certificate Authorities page on the External tab while the external CA list requests (acme, azure-ad-cs, adcs, aws-pca, digicert, aws-acm-public-ca, venafi-tpp, godaddy) answer with a server error.
  3. Look at the External Certificate Authorities panel.
lines
infisical/frontend/src/hooks/api/ca/queries.tsx:95
infisical/frontend/src/hooks/api/ca/queries.tsx:116
infisical/frontend/src/pages/cert-manager/CertificateAuthoritiesPage/components/ExternalCaTable.tsx:51
infisical/frontend/src/pages/cert-manager/CertificateAuthoritiesPage/components/ExternalCaTable.tsx:63
before the PRs testedno
A KMIP client destroys a key re-enabled at the same moment
s15-kmip-destroy-destroys-re-enabled-key-1
kmsbehaviourloss: work
medium

When You re-enable a key just as a KMIP client destroys it.

Got Both say they succeeded and the key is gone.

Want The re-enabled key is kept, because an active key cannot be destroyed.

steps
  1. Create a KMS project with a KMIP client allowed to destroy keys, issue its client certificate, and register an organisation KMIP server certificate.
  2. Add a customer-managed key to the project and disable it.
  3. In turn, re-enable the key then destroy it through KMIP — the destroy is refused because the key is active.
  4. On a fresh disabled key, re-enable it and destroy it through KMIP at the same moment.
  5. Read the key back.
a customer-managed key re-enabled while a KMIP client destroys it is kept
e2e/specs/kms/s15-kmip-destroy-destroys-re-enabled-key.spec.ts:26
lines
infisical/backend/src/ee/services/kmip/kmip-operation-service.ts:151
infisical/backend/src/ee/services/kmip/kmip-operation-service.ts:145
infisical/backend/src/ee/services/kmip/kmip-operation-service.ts:135
infisical/backend/src/services/cmek/cmek-service.ts:116
before the PRs testedno
Opening a certificate authority reads every CA's configuration on the instance
s6-ca-lookup-reads-every-ca-configuration-1
certificate-authoritiesperformanceloss: time
medium

When You open a certificate authority's details, or any action on one CA.

Got The product looks through the configuration of every CA on the instance to find this one's, so it slows as the instance grows.

Want It finds this CA's configuration directly.

steps
  1. Sign in as an admin of a Certificate Management project that holds an internal root CA, on an instance where other organisations have their own CAs
  2. Open Certificate Authorities and open the CA's details page (the app reads the CA by id)
  3. Read the indexes of the internal and external certificate authority tables in Postgres
s6-ca-lookup-reads-every-ca-configuration: ${table}
e2e/specs/certificate-authorities/s6-ca-lookup-reads-every-ca-configuration.spec.ts:10
lines
infisical/backend/src/db/migrations/20250521110635_add-external-ca-pki.ts:35
infisical/backend/src/db/migrations/20250521110635_add-external-ca-pki.ts:88
infisical/backend/src/services/certificate-authority/certificate-authority-dal.ts:120
infisical/backend/src/services/certificate-authority/certificate-authority-dal.ts:122
infisical/backend/src/services/certificate-authority/certificate-authority-dal.ts:127
before the PRs testedno
Listing a project's certificate authorities reads every project's CAs
s6-ca-list-scans-every-project-1
certificate-authoritiesperformanceloss: time
low

When You open a project's Certificate Authorities page.

Got The product looks through the certificate authorities of every project on the instance to find yours.

Want It finds your project's certificate authorities directly, however many others exist.

steps
  1. Sign in as a project admin of a Certificate Management project
  2. Open the project's Certificate Authorities page, which lists its internal CAs
  3. Read the indexes of the certificate_authorities table in the store
listing a project's certificate authorities is served by a projectId index
e2e/specs/certificate-authorities/s6-ca-list-scans-every-project.spec.ts:10
lines
infisical/backend/src/services/certificate-authority/certificate-authority-service.ts:719
infisical/backend/src/services/certificate-authority/certificate-authority-dal.ts:274
infisical/backend/src/db/migrations/20240614154212_certificate-mgmt.ts:23
infisical/backend/src/db/migrations/20250521110635_add-external-ca-pki.ts:76
before the PRs testedno
Two overlapping CA renewals can leave the shorter-lived certificate active
s6-overlapping-ca-renewals-keep-shorter-cert-1
certificate-authoritiesbehaviourloss: decision
medium

When Two people renew the same root CA at about the same time, one to a later expiry and one to an earlier one.

Got Both renewals succeed. The CA can end up on the certificate that expires sooner, and two certificates share one version number.

Want The shorter renewal is refused, as it is when the two are done one after the other.

steps
  1. Create an active internal root CA expiring 2035-01-01 in a Certificate Management project.
  2. Control: renew it to 2045, then renew it to 2040 — the second is refused with 'New Root CA certificate must have notAfter date that is greater than the current certificate notAfter date'.
  3. Create a second such root CA.
  4. Start a renewal of the second CA to 2045 and, while it is still being written (forced here by holding the CA's internal_certificate_authorities row so its final update waits), send a renewal of the same CA to 2040.
  5. Let both finish, then read the CA and its CA certificates list.
a shorter renewal arriving during a longer one is refused and versions stay distinct
e2e/specs/certificate-authorities/s6-overlapping-ca-renewals-keep-shorter-cert.spec.ts:28
lines
infisical/backend/src/services/certificate-authority/internal/internal-certificate-authority-service.ts:880
infisical/backend/src/services/certificate-authority/internal/internal-certificate-authority-service.ts:907
infisical/backend/src/services/certificate-authority/internal/internal-certificate-authority-service.ts:946
infisical/backend/src/services/certificate-authority/internal/internal-certificate-authority-service.ts:990
infisical/backend/src/services/certificate-authority/internal/internal-certificate-authority-service.ts:998
infisical/backend/src/db/migrations/20240802181855_ca-cert-version.ts:27
before the PRs testedno
An active CA's details page says it has no CA certificates when they fail to load
s6-ca-certificates-failure-shown-as-none-1
certificate-authoritiesbehaviourloss: decision
medium

When You open an active certificate authority and its certificates can't be loaded.

Got The page says the CA has no CA certificates installed and invites you to install one.

Want The page says the certificates could not be loaded.

steps
  1. Sign in and open a Certificate Management project.
  2. Create an active internal root CA (with an expiry date, so its certificate is generated).
  3. Make the CA certificates request for that CA fail with a server error.
  4. Open the CA's details page and look at the CA Certificates section.
a failed load of a CA's certificates is reported, not shown as none installed
e2e/specs/certificate-authorities/s6-ca-certificates-failure-shown-as-none.spec.ts:14
lines
infisical/frontend/src/pages/cert-manager/CertAuthDetailsByIDPage/components/CaCertificatesSection/CaCertificatesTable.tsx:36
infisical/frontend/src/pages/cert-manager/CertAuthDetailsByIDPage/components/CaCertificatesSection/CaCertificatesTable.tsx:51
infisical/frontend/src/hooks/api/ca/queries.tsx:166
before the PRs testedno
A failed CRL load on a certificate authority's page says "This CA does not have any CRLs" instead of an error
s3-ca-crls-failure-shown-as-none-1
behaviourloss: decision
medium

When

Got

Want

steps
  1. Create an active internal root CA in a Certificate Management project (it gets a CRL at creation).
  2. Open the CA's details page while the CRL list request answers with a server error.
  3. Look at the CA Certificate Revocation Lists (CRLs) section.
lines
infisical/frontend/src/pages/cert-manager/CertAuthDetailsByIDPage/components/CaCrlsSection/CaCrlsTable.tsx:28
infisical/frontend/src/pages/cert-manager/CertAuthDetailsByIDPage/components/CaCrlsSection/CaCrlsTable.tsx:43
infisical/frontend/src/hooks/api/ca/queries.tsx:214
before the PRs testedno
Listing a CA's certificate versions slows with every renewal on the instance
s7-ca-certificate-history-unindexed-1
certificate-authoritiesperformanceloss: time
medium

When You open a certificate authority's details page.

Got Listing its certificate versions gets slower as the instance keeps more CA certificates and renewals, across every organisation.

Want The versions load at the same cost however many CA certificates the instance keeps.

steps
  1. Sign in and open a Certificate Management project.
  2. Create an active internal root CA.
  3. Open the CA's details page, which lists its CA certificate versions (GET /api/v1/cert-manager/ca/internal/:caId/ca-certificates).
  4. Look at the indexes of the certificate_authority_certs table in the database.
a CA's certificates are read through an index on the CA
e2e/specs/certificate-authorities/s7-ca-certificate-history-unindexed.spec.ts:9
lines
infisical/backend/src/services/certificate-authority/certificate-authority-fns.ts:518
infisical/backend/src/services/certificate-authority/internal/internal-certificate-authority-service.ts:1200
infisical/backend/src/db/migrations/20240802181855_ca-cert-version.ts:28
before the PRs testedno
Closing the Share a Secret panel throws away the typed secret value without a warning
s7-share-dialog-close-wipes-typed-secret-1
behaviourloss: work
low

When

Got

Want

steps
  1. Sign in and open Secret Sharing for your organisation
  2. Click Share Secret
  3. Type a value in Secret Value
  4. Close the panel with its Close (X) button
  5. Click Share Secret again
lines
infisical/frontend/src/pages/organization/SecretSharingPage/components/ShareSecret/AddShareSecretModal.tsx:19
infisical/frontend/src/pages/public/ShareSecretPage/components/ShareSecretForm.tsx:202
before the PRs testedno
A certificate authority's details page goes blank, with no error, when the CA fails to load or no longer exists
s2-ca-details-blank-when-ca-fails-to-load-1
uiloss: time
medium

When

Got

Want

steps
  1. Sign in as a user whose organisation has a Certificate Management project holding an internal root CA
  2. Open that CA's details page (Certificate Authorities, then the CA's name)
  3. Have the CA's own load fail with a server error (both the first request and the automatic retry), or open the details page of a CA id that does not exist
  4. Look at the page content area
lines
infisical/frontend/src/pages/cert-manager/CertAuthDetailsByIDPage/CertAuthDetailsByIDPage.tsx:61
infisical/frontend/src/pages/cert-manager/CertAuthDetailsByIDPage/CertAuthDetailsByIDPage.tsx:105
infisical/frontend/src/pages/cert-manager/CertAuthDetailsByIDPage/components/CaDetailsSection.tsx:169
infisical/frontend/src/pages/cert-manager/CertAuthDetailsByIDPage/components/CaCertDetailsSection.tsx:27
before the PRs testedno
Installing an intermediate CA's certificate can report an error for an install that succeeded
s8-concurrent-intermediate-installs-one-refused-after-applied-1
certificate-authoritiesbehaviourloss: time
low

When You install an intermediate CA's certificate twice at nearly the same moment, for example from two tabs.

Got One install says something went wrong, but its certificate was installed anyway and the CA now holds two.

Want An install that reports a failure installs nothing, or both installs succeed.

steps
  1. In a Certificate Management project, create an active internal root CA and a pending intermediate CA
  2. Send two installs of the intermediate's certificate with the root as parent at the same moment
  3. Look at the two answers and at the intermediate's CA certificates
an intermediate CA install answered with an error leaves no certificate installed
e2e/specs/certificate-authorities/s8-concurrent-intermediate-installs-one-refused-after-applied.spec.ts:54
lines
infisical/backend/src/services/certificate-authority/internal/internal-certificate-authority-service.ts:1856
infisical/backend/src/services/certificate-authority/internal/internal-certificate-authority-service.ts:1865
infisical/backend/src/services/certificate-authority/internal/internal-certificate-authority-service.ts:1867
infisical/backend/src/db/migrations/20260218175924_add-ca-signing-configs.ts:11
before the PRs testedno
The certificate Inventory says a project has no certificates when its certificate list fails to load
s3-failed-certificate-inventory-shown-as-no-certificates-1
behaviourloss: decision
medium

When

Got

Want

steps
  1. Sign in and open a Certificate Manager project.
  2. Make the certificate list request fail with a server error (POST /api/v1/projects/<projectId>/certificates/search answers 500).
  3. Open the project's Inventory.
  4. Wait for the read to give up (it is retried once).
lines
infisical/frontend/src/pages/cert-manager/CertificatesPage/components/CertificatesTable.tsx:380
infisical/frontend/src/pages/cert-manager/CertificatesPage/components/CertificatesTable.tsx:1389
infisical/frontend/src/hooks/api/projects/queries.tsx:611
before the PRs testedno
Certificate Policies hides every policy older than the newest 100, and the Create Profile wizard cannot pick them
s4-certificate-policies-beyond-100-unreachable-1
behaviourloss: time
medium

When

Got

Want

steps
  1. Create a Certificate Manager project
  2. Create 101 certificate policies in it
  3. Open the project's Certificate Policies
  4. Scroll to the end of the list and look for a pager or a way to find the oldest policy
lines
infisical/frontend/src/pages/cert-manager/PoliciesPage/components/CertificatePoliciesTab/PolicyList.tsx:55
infisical/frontend/src/pages/cert-manager/PoliciesPage/components/CertificateProfilesTab/CreateProfileModal.tsx:568
infisical/frontend/src/pages/cert-manager/PoliciesPage/components/CertificateProfilesTab/ProfileList.tsx:30
infisical/backend/src/server/routes/v1/certificate-policy-router.ts:261
infisical/backend/src/services/certificate-policy/certificate-policy-dal.ts:183
before the PRs testedno
Certificate Profiles says a project has no profiles when the list fails to load
s9-failed-certificate-profiles-shown-as-none-1
certificatesbehaviourloss: decision
medium

When You open Certificate Profiles while the profile list can't be loaded.

Got The page says there are no certificate profiles and invites you to create your first one. Nothing says the load failed.

Want The page says the profiles could not be loaded.

steps
  1. Sign in and open a Certificate Manager project.
  2. Make the certificate profile list request fail with a server error (GET /api/v1/cert-manager/certificate-profiles answers 500).
  3. Open the project's Certificate Profiles page.
  4. Wait for the read to give up (it is retried once).
s9-failed-certificate-profiles-shown-as-none
e2e/specs/certificates/s9-failed-certificate-profiles-shown-as-none.spec.ts:10
lines
infisical/frontend/src/pages/cert-manager/PoliciesPage/components/CertificateProfilesTab/ProfileList.tsx:30
infisical/frontend/src/pages/cert-manager/PoliciesPage/components/CertificateProfilesTab/ProfileList.tsx:58
infisical/frontend/src/hooks/api/certificateProfiles/queries.tsx:59
before the PRs testedno
Certificate Policies says a project has none when its policies fail to load
s9-failed-certificate-policies-shown-as-none-1
certificatesbehaviourloss: decision
medium

When You open Certificate Policies and the policies can't be loaded.

Got The screen says the project has no certificate policies and invites you to create one.

Want It says the policies could not be loaded.

steps
  1. Sign in and open a Certificate Manager project that holds a certificate policy.
  2. Make the policy list request fail with a server error (GET /api/v1/cert-manager/certificate-policies answers 500).
  3. Open the project's Certificate Policies screen.
  4. Wait for the read to give up (it is retried once).
s9-failed-certificate-policies-shown-as-none
e2e/specs/certificates/s9-failed-certificate-policies-shown-as-none.spec.ts:11
lines
infisical/frontend/src/pages/cert-manager/PoliciesPage/components/CertificatePoliciesTab/PolicyList.tsx:55
infisical/frontend/src/pages/cert-manager/PoliciesPage/components/CertificatePoliciesTab/PolicyList.tsx:93
before the PRs testedno
Opening several shared secrets at once fails some of them with a server error
s9-concurrent-share-access-exhausts-pool-1
secret-sharingperformanceloss: time
medium

When You open several shared secrets at the same moment.

Got Some fail with a generic error after a long wait, and you have to try again.

Want Every one opens, or you are clearly told it is busy.

steps
  1. Sign in as a member and create 20 shared secrets, each password-protected and restricted to your own email, via POST /api/v1/shared-secrets.
  2. Open all 20 at the same moment with the correct password, via POST /api/v1/shared-secrets/:id/access.
  3. Observe that about half (10-11 of 20) return 500 'Something went wrong' after a ~60 s stall, while the rest return 200.
lines
infisical/backend/src/services/secret-sharing/secret-sharing-service.ts:689
infisical/backend/src/services/secret-sharing/secret-sharing-service.ts:717
infisical/backend/src/services/secret-sharing/secret-sharing-service.ts:609
infisical/backend/src/services/secret-sharing/secret-sharing-service.ts:770
before the PRs testedno
Export Certificate (PEM) silently shows no private key when the certificate's key bundle fails to load
s5-pem-export-failed-bundle-drops-private-key-1
behaviourloss: time
medium

When

Got

Want

steps
  1. Sign in as a user allowed to read certificate private keys in a cert-manager project
  2. Import a certificate together with its private key and chain
  3. Open the certificate's details page
  4. Make the certificate bundle read fail on the server (GET /api/v1/pki/certificates/<serial>/bundle answers 500)
  5. Choose Options → Export Certificate → Export PEM
  6. Observe the Export Certificate dialog
lines
infisical/frontend/src/pages/cert-manager/CertificatesPage/components/CertificateCertModal.tsx:56
infisical/frontend/src/pages/cert-manager/CertificatesPage/components/CertificateCertModal.tsx:69
infisical/frontend/src/pages/cert-manager/CertificatesPage/components/CertificateCertModal.tsx:93
before the PRs testedno
An audit log filter that matches nothing says "No audit logs on file", as if the organization had no logs
s4-no-match-audit-log-filter-shown-as-no-logs-1
uiloss: decision
low

When

Got

Want

steps
  1. Sign in as the admin of an organization that has recorded audit logs.
  2. Open the organization's Audit Logs page and wait for the events to list.
  3. In the search box add the filter actor_id:00000000-0000-0000-0000-000000000000 and press Enter.
  4. Look at the table once the filtered load finishes.
lines
infisical/frontend/src/pages/organization/AuditLogsPage/components/LogsTable.tsx:56
infisical/frontend/src/pages/organization/AuditLogsPage/components/LogsTable.tsx:67
infisical/frontend/src/pages/project/AuditLogsPage/components/ProjectAuditLogsTabs.tsx:50
before the PRs testedno
Cancelling a new log stream throws away the typed endpoint URL without asking
s9-log-stream-draft-discarded-on-cancel-1
audit-logsuiloss: time
low

When You start a Custom log stream, type its endpoint URL, then press Cancel.

Got The panel closes straight away and the URL you typed is gone when you open Custom again.

Want You are asked whether to discard what you typed, or the URL is kept.

steps
  1. Sign in as an organization admin and open Audit Logs, External Log Streams tab
  2. Click Add Log Stream and pick Custom
  3. Type an Endpoint URL
  4. Click Cancel
  5. Click Add Log Stream and pick Custom again
s9-log-stream-draft-discarded-on-cancel
e2e/specs/audit-logs/s9-log-stream-draft-discarded-on-cancel.spec.ts:6
lines
infisical/frontend/src/pages/organization/AuditLogsPage/components/AuditLogStreamTab/components/AddAuditLogStreamModal.tsx:19
infisical/frontend/src/pages/organization/AuditLogsPage/components/AuditLogStreamTab/AuditLogStreamForm/AuditLogStreamFormFooter.tsx:40
before the PRs testedno
The Certificate Profiles screen sends one extra request for every profile listed
s10-profile-policies-fetched-one-request-per-row-1
certificatesperformanceloss: time
medium

When You open a project's Certificate Profiles screen.

Got The screen asks the server for each profile's policy separately, so it does more work the more profiles you have.

Want The screen gets every policy name in the one profile list it already loads.

steps
  1. Sign in as a user with a Certificate Manager project
  2. Create 3 certificate policies and 3 self-signed certificate profiles, each bound to its own policy, in one project
  3. Create 12 such policies and profiles in a second project
  4. Open each project's Certificate Profiles screen and count the requests for a single certificate policy while it loads
s10-profile-policies-fetched-one-request-per-row
e2e/specs/certificates/s10-profile-policies-fetched-one-request-per-row.spec.ts:42
lines
infisical/frontend/src/pages/cert-manager/PoliciesPage/components/CertificateProfilesTab/ProfileRow.tsx:66
infisical/frontend/src/pages/cert-manager/PoliciesPage/components/CertificateProfilesTab/ProfileRow.tsx:50
infisical/frontend/src/pages/cert-manager/PoliciesPage/components/CertificateProfilesTab/ProfileList.tsx:31
infisical/backend/src/services/certificate-profile/certificate-profile-dal.ts:381
before the PRs testedno
The Native Integrations search is cleared after opening a found integration and going back to the list
s6-native-integrations-search-lost-on-back-1
uiloss: time
low

When

Got

Want

steps
  1. Open a Secrets Management project that has a native AWS Parameter Store integration and go to Integrations → Native Integrations
  2. Type the integration's secret path into "Search integrations..." so the list shows only that integration
  3. Click the integration's row to open its detail page
  4. Press the browser's Back button
  5. Observe the search box is empty and the list shows every integration again
lines
infisical/frontend/src/pages/secret-manager/IntegrationsListPage/components/NativeIntegrationsTab/IntegrationsTable.tsx:124
infisical/frontend/src/hooks/usePagination.tsx:20
infisical/frontend/src/pages/secret-manager/IntegrationsListPage/components/NativeIntegrationsTab/IntegrationRow.tsx:76
before the PRs testedno
Two revocations of one certificate at once both succeed, and the second rewrites the reason
s11-overlapping-revocations-both-accepted-1
certificatesbehaviourloss: work
medium

When Two people revoke the same certificate at the same moment, with different reasons.

Got Both are told the revocation worked. The certificate keeps the second reason and time, and the first person's reason (for example key compromise) is gone for good.

Want One revocation wins and the other is told the certificate is already revoked.

steps
  1. Create a Certificate Manager project with an active internal root CA, a certificate policy and an API-enrollment certificate profile on it
  2. Issue a certificate from the profile, revoke it with reason KEY_COMPROMISE, then revoke it again with SUPERSEDED: the second is refused
  3. Issue a fresh certificate and hold its row so both writes meet (a row lock on certificates where id = the new certificate)
  4. Send two revocations of the fresh certificate: KEY_COMPROMISE and SUPERSEDED
  5. Release the hold and read both answers and the certificate
two overlapping revocations of one certificate: one is accepted, the other refused
e2e/specs/certificates/s11-overlapping-revocations-both-accepted.spec.ts:16
lines
infisical/backend/src/services/certificate/certificate-service.ts:628
infisical/backend/src/services/certificate/certificate-service.ts:714
infisical/backend/src/services/certificate/certificate-service.ts:758
infisical/backend/src/services/certificate/certificate-service.ts:771
before the PRs testedno
Renewing one certificate twice at once issues two replacement certificates
s11-overlapping-renewals-issue-two-replacements-1
certificatesbehaviourloss: time
low

When A certificate is renewed twice at nearly the same moment, for example by you and the automatic renewal.

Got Two new certificates are issued, though a second renewal later is refused as already renewed.

Want Only one replacement is issued and the other renewal is refused.

steps
  1. Create a Certificate Manager project, an active internal root CA with a notAfter date, a certificate policy and an API-enrollment certificate profile on that CA
  2. Issue a certificate from the profile
  3. Renew it, then renew the same original again: the second renewal is refused with "Certificate has already been renewed"
  4. Issue a fresh certificate from the profile
  5. Send two renewals of the fresh certificate while the first is still being written (forced by holding the original certificate's row so both reach their final write)
  6. Count the certificates recorded as renewed from the fresh original
overlapping renewals of one certificate issue one replacement
e2e/specs/certificates/s11-overlapping-renewals-issue-two-replacements.spec.ts:20
lines
infisical/backend/src/services/certificate-v3/certificate-renewal-service.ts:590
infisical/backend/src/services/certificate-v3/certificate-renewal-fns.ts:702
infisical/backend/src/services/certificate-v3/certificate-renewal-service.ts:286
before the PRs testedno
"Select another provider" on the Add Log Stream form throws away the typed endpoint URL
s4-log-stream-draft-lost-on-provider-switch-1
behaviourloss: time
low

When

Got

Want

steps
  1. Sign in as an organization admin and open Audit Logs > External Log Streams.
  2. Click Add Log Stream, then pick Custom.
  3. Type an Endpoint URL.
  4. Click "Select another provider" at the top of the sheet.
  5. Pick Custom again.
lines
infisical/frontend/src/pages/organization/AuditLogsPage/components/AuditLogStreamTab/components/AddAuditLogStreamModal.tsx:38
infisical/frontend/src/pages/organization/AuditLogsPage/components/AuditLogStreamTab/components/AddAuditLogStreamModal.tsx:67
before the PRs testedno
A certificate assigned to two Applications at once is accepted for both
s11-certificate-assigned-to-two-applications-at-once-1
certificatesbehaviourloss: decision
medium

When Two admins attach the same new certificate to different Applications at the same moment.

Got Both are told it worked, but the certificate ends up in only one Application.

Want One is attached and the other admin is told the certificate is already assigned.

steps
  1. Sign up a new organization; its Certificate Manager project is the active one
  2. Create two Applications, A and B, in that project
  3. Import a certificate and assign it to A, then to B: the second is refused as 'already assigned to an Application and cannot be moved'
  4. Import a second certificate
  5. Assign it to A and to B at the same moment (forced in the test by holding the certificate row so both requests reach their write)
  6. Read both answers and the certificate's stored Application
s11-certificate-assigned-to-two-applications-at-once
e2e/specs/certificates/s11-certificate-assigned-to-two-applications-at-once.spec.ts:44
lines
infisical/backend/src/services/certificate/certificate-service.ts:1720
infisical/backend/src/services/certificate/certificate-service.ts:1725
infisical/backend/src/services/certificate/certificate-service.ts:1750
before the PRs testedno
The certificate profiles list slows down with every EST profile a project adds
s10-est-profile-list-reads-config-per-profile-1
certificatesperformanceloss: time
medium

When A Certificate Manager project has many EST-enrolled certificate profiles.

Got Opening Certificate Profiles does extra work for each EST profile, so the list gets slower as profiles are added.

Want The list loads in the same amount of work however many EST profiles the project has.

steps
  1. Sign in as a user of a Certificate Manager project with an active internal root CA and a certificate policy.
  2. Create 2 certificate profiles with EST enrollment, issued by that CA, each with a passphrase and a bootstrap CA chain.
  3. List the project's certificate profiles (limit 100) and count the database statements the list makes.
  4. Create 10 more such EST profiles, so the project holds 12.
  5. List the profiles again and count the statements.
s10-est-profile-list-reads-config-per-profile
e2e/specs/certificates/s10-est-profile-list-reads-config-per-profile.spec.ts:21
lines
infisical/backend/src/services/certificate-profile/certificate-profile-service.ts:1350
infisical/backend/src/services/certificate-profile/certificate-profile-service.ts:1357
infisical/backend/src/services/certificate-profile/certificate-profile-service.ts:1359
infisical/backend/src/services/certificate-profile/certificate-profile-service.ts:295
infisical/backend/src/services/certificate-profile/certificate-profile-dal.ts:446
before the PRs testedno
A profile deleted during an attach removes an attachment reported as done
s12-profile-deleted-during-attach-drops-attachment-1
certificatesbehaviourloss: decision
low

When One admin attaches a certificate profile to an Application while another deletes that profile.

Got Both are told it worked. The profile is gone and the Application no longer has it attached.

Want One of the two is told it could not be done.

steps
  1. Create an internal root CA, a certificate policy, an API-enrollment certificate profile and an Application in an organization's active Certificate Manager project.
  2. Control: attach a second profile to the Application, then delete it — the delete is refused because the profile is attached.
  3. Hold the new profile's row in the database so writes to it wait.
  4. Attach the new profile to the Application; it waits on the profile's foreign-key check.
  5. Delete the new profile; its attachment read finds nothing and the delete waits on the row.
  6. Release the hold and read both answers, then list the Application's profiles.
a profile deleted while being attached is either refused or keeps its attachment
e2e/specs/certificates/s12-profile-deleted-during-attach-drops-attachment.spec.ts:41
lines
infisical/backend/src/services/certificate-profile/certificate-profile-service.ts:1453
infisical/backend/src/services/certificate-profile/certificate-profile-service.ts:1467
infisical/backend/src/db/migrations/20260504432352_cert-manager-applications.ts:106
infisical/backend/src/services/pki-application/pki-application-service.ts:499
before the PRs testedno
The Applications page says a Certificate Management project has no applications when its application list fails to load
s2-applications-load-failure-shown-as-empty-1
behaviourloss: decision
medium

When

Got

Want

steps
  1. Sign in as a project admin and open a Certificate Management project.
  2. Make the applications list request fail with a server error (GET /api/v1/cert-manager/applications answers 500).
  3. Open the project's Applications page.
  4. Wait for the read to give up (it is retried once).
lines
infisical/frontend/src/pages/cert-manager/ApplicationsPage/ApplicationsPage.tsx:63
infisical/frontend/src/pages/cert-manager/ApplicationsPage/ApplicationsPage.tsx:85
infisical/frontend/src/hooks/api/pkiApplications/queries.ts:37
before the PRs testedno
Closing the Add Reminder form throws away the typed note without asking
s7-reminder-note-discarded-on-close-1
secret-remindersuiloss: time
low

When You start adding a reminder to a secret, type a note, then press Close.

Got The form closes straight away and your note is gone when you open it again.

Want You are asked whether to discard what you typed, or your note is kept.

steps
  1. Sign in and open a secret-management project's secrets overview
  2. Open a secret's annotations and choose Add Reminder
  3. Type a note in the reminder form
  4. Press Close in the form's footer
  5. Open Add Reminder on the same secret again
s7-reminder-note-discarded-on-close
e2e/specs/secret-reminders/s7-reminder-note-discarded-on-close.spec.ts:5
lines
infisical/frontend/src/pages/secret-manager/OverviewPage/components/SecretTableRow/SecretReminderForm.tsx:487
infisical/frontend/src/pages/secret-manager/OverviewPage/components/SecretTableRow/SecretEditTableRow.tsx:1397
infisical/frontend/src/pages/secret-manager/OverviewPage/components/SecretTableRow/SecretEditTableRow.tsx:1413
infisical/frontend/src/pages/secret-manager/OverviewPage/components/CreateDynamicSecretForm/CreateDynamicSecretForm.tsx:61
before the PRs testedno
Starring a project after favourites fail to load erases your other favourites
s6-favourites-load-failure-wipes-favourites-1
behaviourloss: work
medium

When

Got

Want

steps
  1. Sign in as a member with one Secrets Management project already starred as a favourite and a second project not starred
  2. Make the request that loads the member's favourites fail with a server error (the load and its automatic retry)
  3. Open the Secrets Management projects list and switch to list view
  4. Find the second project and choose Add to favorites
  5. Reload the page once the favourites load works again
lines
infisical/frontend/src/pages/organization/ProjectsPage/ProjectTypePage/ProjectTypePage.tsx:458
infisical/frontend/src/pages/organization/ProjectsPage/ProjectTypePage/ProjectTypePage.tsx:483
infisical/frontend/src/pages/organization/ProjectsPage/ProjectTypePage/ProjectTypePage.tsx:494
infisical/frontend/src/hooks/api/users/mutation.tsx:55
infisical/backend/src/services/user/user-service.ts:820
before the PRs testedno
Starring two projects quickly keeps only the second star
s5-second-favorite-star-drops-first-1
behaviourloss: time
low

When

Got

Want

steps
  1. Open the Secrets Management projects list in list view, with two of your own projects not yet starred, on a slow connection
  2. Star the first project
  3. Before the favourites list refreshes, star the second project
  4. Reload the page and look at the first project
lines
infisical/frontend/src/pages/organization/ProjectsPage/ProjectTypePage/ProjectTypePage.tsx:494
infisical/frontend/src/hooks/api/users/mutation.tsx:63
infisical/backend/src/services/user/user-service.ts:851
before the PRs testedno
An application's profiles look detached when they fail to load
s8-application-profiles-load-failure-shown-as-none-1
pki-enrollmentuiloss: time
medium

When You open an application's settings and its profiles can't be loaded.

Got It says no profiles are attached, even though one is, and nothing tells you something went wrong.

Want It tells you the profiles could not be loaded.

steps
  1. Create a Certificate Management project with an active internal CA, a certificate profile, and a PKI application; attach the profile to the application.
  2. Open the project's Applications page and open the application.
  3. While the application's attached-profiles list fails to load (server error), open its Settings tab.
s8-application-profiles-load-failure-shown-as-none
e2e/specs/pki-enrollment/s8-application-profiles-load-failure-shown-as-none.spec.ts:13
lines
infisical/frontend/src/pages/cert-manager/ApplicationDetailsByIDPage/ApplicationDetailsByIDPage.tsx:100
infisical/frontend/src/pages/cert-manager/ApplicationDetailsByIDPage/components/ApplicationSettingsTab.tsx:462
infisical/frontend/src/hooks/api/pkiApplications/queries.ts:116
infisical/frontend/src/pages/cert-manager/ApplicationsPage/components/ConfigureProfilesModal.tsx:36
before the PRs testedno
A reminder saved while it is removed stays set with no one to remind
s8-reminder-removed-while-saved-keeps-no-recipients-1
secret-remindersbehaviourloss: work
low

When One person saves a secret's reminder at the same moment another removes it.

Got The reminder stays on the secret, but it has no recipients left.

Want The reminder is either removed, or kept with the people it was saved for.

steps
  1. Create a secret-management project and a secret in it.
  2. Set a reminder on the secret (repeat every day) with yourself as recipient.
  3. Start saving the reminder again with a new note, and while that save waits on the reminder's alert, delete the reminder.
  4. Let both finish (both answer OK).
  5. Read the secret's reminder back.
a reminder saved while it is removed keeps its recipients or is gone
e2e/specs/secret-reminders/s8-reminder-removed-while-saved-keeps-no-recipients.spec.ts:11
lines
infisical/backend/src/services/reminder/reminder-service.ts:345
infisical/backend/src/services/reminder/reminder-service.ts:346
infisical/backend/src/services/reminder/reminder-service.ts:226
infisical/backend/src/services/reminder/reminder-service.ts:231
infisical/backend/src/services/alert/alert-service.ts:491
infisical/backend/src/services/alert/alert-service.ts:713
before the PRs testedno
"Mark all as read" rewrites every notification, not just the unread ones
s9-mark-all-read-rewrites-read-notifications-1
secret-remindersperformanceloss: time
low

When You click Mark all as read with one new notification and a long history of read ones.

Got Every notification you have is saved again, so the action does more work the longer your history.

Want Only the unread notification is changed.

steps
  1. Sign up a fresh account and its organization
  2. Sign in from 5 different devices so the product writes 5 new-device notifications, then mark all as read
  3. Sign in from one more new device so exactly one notification is unread
  4. Mark all as read and count the rows the update wrote (pg_stat_statements rows for the user_notifications isRead update)
  5. Repeat with 30 more read notifications and one unread
s9 mark all as read rewrites only the unread notifications
e2e/specs/secret-reminders/s9-mark-all-read-rewrites-read-notifications.spec.ts:29
lines
infisical/backend/src/services/notification/user-notification-dal.ts:119
infisical/backend/src/services/notification/user-notification-dal.ts:125
infisical/backend/src/services/notification/notification-service.ts:48
before the PRs testedno
The notification bell downloads every recent notification every 30 seconds
s7-bell-loads-every-notification-1
secret-remindersperformanceloss: time
medium

When You have collected many notifications over the last three months and keep an organization page open.

Got Every 30 seconds the page downloads all of them again, with their full text, though the bell panel shows 20 at a time.

Want The page fetches only the 20 the panel shows, and the next 20 when you ask for more.

steps
  1. Sign up a throwaway account with its own organization.
  2. Sign in five times, each from a different browser (a new User-Agent), so the product writes five new-device notifications.
  3. Fetch the bell's notification list and note the count and size.
  4. Sign in thirty more times from new devices.
  5. Fetch the bell's notification list again.
s7-bell-loads-every-notification
e2e/specs/secret-reminders/s7-bell-loads-every-notification.spec.ts:22
lines
infisical/backend/src/services/notification/user-notification-dal.ts:24
infisical/backend/src/services/notification/user-notification-dal.ts:46
infisical/backend/src/services/notification/notification-service.ts:23
infisical/backend/src/server/routes/v1/notification-router.ts:32
infisical/frontend/src/hooks/api/notifications/queries.tsx:29
infisical/frontend/src/layouts/OrganizationLayout/components/NavBar/NotificationDropdown.tsx:33
infisical/frontend/src/layouts/OrganizationLayout/components/NavBar/NotificationDropdown.tsx:75
before the PRs testedno
A profile's enrollment setup looks empty when it fails to load
s8-enrollment-load-failure-shown-as-unconfigured-1
pki-enrollmentbehaviourloss: decision
medium

When You open the enrollment setup of an application's profile and it cannot be loaded.

Got The dialog says no enrollment methods are configured and invites you to add one, although the profile already has API enrollment.

Want The dialog says the enrollment setup could not be loaded.

steps
  1. Sign in as a project admin of a Certificate Management project with an active internal CA.
  2. Create a certificate profile, an application, attach the profile to the application and enable API enrollment for it.
  3. Open the application and its Settings tab; the profile row shows the API enrollment badge.
  4. Make the enrollment configuration request fail with a server error.
  5. Click the profile's API enrollment badge to open its enrollment configuration.
s8-enrollment-load-failure-shown-as-unconfigured
e2e/specs/pki-enrollment/s8-enrollment-load-failure-shown-as-unconfigured.spec.ts:13
lines
infisical/frontend/src/pages/cert-manager/ApplicationDetailsByIDPage/components/ConfigureEnrollmentModal.tsx:1396
infisical/frontend/src/pages/cert-manager/ApplicationDetailsByIDPage/components/ConfigureEnrollmentModal.tsx:1516
infisical/frontend/src/pages/cert-manager/ApplicationDetailsByIDPage/components/ConfigureEnrollmentModal.tsx:1569
before the PRs testedno
Screen readers announce the reminder recipient's remove button as "Remove [object Object]"
s9-reminder-recipient-remove-unnamed-1
secret-remindersuiloss: time
low

When You open Add Reminder on a secret with a screen reader on.

Got The button that removes the All Project Members recipient is read out as "Remove object Object".

Want It is read as "Remove All Project Members".

steps
  1. Sign in and open a secret-management project's secrets overview
  2. Open a secret's annotations and choose Add Reminder
  3. Move a screen reader (or the accessibility tree) to the remove control on the All Project Members recipient
s9-reminder-recipient-remove-unnamed
e2e/specs/secret-reminders/s9-reminder-recipient-remove-unnamed.spec.ts:5
lines
infisical/frontend/src/views/Alerts/ChannelRecipientsField.tsx:49
infisical/frontend/src/views/Alerts/ChannelRecipientsField.tsx:105
infisical/frontend/src/components/v3/generic/ReactSelect/components.tsx:27
before the PRs testedno
A PKI subscriber's page is blank, with no error, when the subscriber fails to load or no longer exists
s3-subscriber-detail-load-failure-blank-1
uiloss: time
medium

When

Got

Want

steps
  1. Sign in as a user whose organisation has a Certificate Management project with a PKI subscriber
  2. Open the subscriber page of a subscriber name that does not exist in that project (for example a saved link to a subscriber since deleted)
  3. Alternatively, open an existing subscriber's page while its record read answers a server error
  4. Look at the page content area
lines
infisical/frontend/src/pages/cert-manager/PkiSubscriberDetailsByIDPage/PkiSubscriberDetailsByIDPage.tsx:45
infisical/frontend/src/pages/cert-manager/PkiSubscriberDetailsByIDPage/PkiSubscriberDetailsByIDPage.tsx:79
before the PRs testedno
A recurring secret reminder can't be edited once its start date has passed
s2-recurring-reminder-note-edit-blocked-by-start-date-1
behaviourloss: time
medium

When

Got

Want

steps
  1. Have a secret with a recurring reminder (every 30 days, note "a") whose start date is yesterday — or one created over the API with no start date
  2. Open the project's secrets overview
  3. Open that secret's reminder from its row
  4. Change only the Note to "rotated note"
  5. Click Update Reminder
  6. Reopen the reminder, or read it back over the API
lines
infisical/frontend/src/pages/secret-manager/OverviewPage/components/SecretTableRow/SecretReminderForm.tsx:95
infisical/frontend/src/pages/secret-manager/OverviewPage/components/SecretTableRow/SecretReminderForm.tsx:102
infisical/frontend/src/pages/secret-manager/OverviewPage/components/SecretTableRow/SecretReminderForm.tsx:187
infisical/backend/src/services/reminder/reminder-service.ts:146
infisical/backend/src/server/routes/v1/reminder-routers/secret-reminder-router.ts:34
before the PRs testedno
Past the 100th certificate, a subscriber's certificate list says "No certificates have been issued for this subscriber" and the pager disappears
s2-subscriber-certificates-past-offset-100-shown-as-none-1
behaviourloss: decision
medium

When

Got

Want

steps
  1. In a Certificate Manager project with an active internal CA, create a PKI subscriber and issue it 130 certificates
  2. Open the subscriber's detail page (Subscribers > the subscriber) at the default 25 per page
  3. Click Next five times to reach page 6 (offset 125)
  4. The table is replaced by 'No certificates have been issued for this subscriber' and the pager is gone
lines
infisical/backend/src/server/routes/v1/pki-subscriber-router.ts:763
infisical/frontend/src/pages/cert-manager/PkiSubscriberDetailsByIDPage/components/PkiSubscriberCertificatesTable.tsx:56
infisical/frontend/src/pages/cert-manager/PkiSubscriberDetailsByIDPage/components/PkiSubscriberCertificatesTable.tsx:113
infisical/frontend/src/pages/cert-manager/PkiSubscriberDetailsByIDPage/components/PkiSubscriberCertificatesTable.tsx:211
before the PRs testedno
The Subscribers page downloads every subscriber of a project at once
s9-subscribers-list-loads-every-subscriber-1
pki-enrollmentperformanceloss: time
medium

When Your project has many PKI subscribers and you open the Subscribers page.

Got The page downloads every subscriber in the project at once, so it gets slower as the project grows.

Want The page loads one page of subscribers at a time.

steps
  1. Sign in as the admin of a Certificate Management project with an active internal CA
  2. Create 120 PKI subscribers in the project
  3. Open the project's Subscribers page
  4. Watch the subscriber list the page downloads
s9-subscribers-list-loads-every-subscriber
e2e/specs/pki-enrollment/s9-subscribers-list-loads-every-subscriber.spec.ts:11
lines
infisical/backend/src/services/project/project-service.ts:1702
infisical/backend/src/server/routes/v1/project-router.ts:1751
infisical/frontend/src/hooks/api/projects/queries.tsx:793
infisical/frontend/src/pages/cert-manager/PkiSubscribersPage/components/PkiSubscribersTable.tsx:55
before the PRs testedno
Going Back in "Add integration" throws away the typed Slack alias
s5-slack-alias-lost-on-back-1
behaviourloss: time
low

When

Got

Want

steps
  1. Sign in as an organization admin and open Integrations > Workflow Integrations.
  2. Click Add integration, then pick Slack.
  3. Type an Alias.
  4. Click Back.
  5. Pick Slack again.
lines
infisical/frontend/src/pages/organization/SettingsPage/components/OrgWorkflowIntegrationTab/AddWorkflowIntegrationForm.tsx:98
infisical/frontend/src/pages/organization/SettingsPage/components/OrgWorkflowIntegrationTab/SlackIntegrationForm.tsx:46
before the PRs testedno
A subscriber's certificates are found by scanning every certificate in the instance
s10-subscriber-certificates-not-indexed-by-subscriber-1
pki-enrollmentperformanceloss: time
medium

When You open a PKI subscriber's certificates or its latest certificate.

Got The product looks through every certificate ever issued on the instance to find this subscriber's few, so it slows as other teams issue more.

Want Only this subscriber's certificates are looked at.

steps
  1. Create a Certificate Management project with an active internal root CA
  2. Create a PKI subscriber on that CA and issue two certificates from it
  3. Open the subscriber's certificate list (GET /api/v1/pki/subscribers/:subscriberName/certificates)
  4. Read the indexes of the certificates table and the plan of the list statement
s10-subscriber-certificates-not-indexed-by-subscriber
e2e/specs/pki-enrollment/s10-subscriber-certificates-not-indexed-by-subscriber.spec.ts:9
lines
infisical/backend/src/db/migrations/20250508160957_pki-subscriber.ts:30
infisical/backend/src/services/pki-subscriber/pki-subscriber-service.ts:740
infisical/backend/src/services/certificate/certificate-dal.ts:398
infisical/backend/src/services/certificate/certificate-dal.ts:88
before the PRs testedno
Cancelling "Add integration" throws away the typed Slack alias without asking
s7-slack-alias-lost-on-cancel-1
webhooks-notificationsbehaviourloss: time
low

When You start connecting Slack, type an alias, and press Cancel.

Got The dialog closes straight away and the alias is gone when you come back, with no warning.

Want You are asked before your typed alias is discarded.

steps
  1. Sign in as an organization admin and open Integrations > Workflow Integrations.
  2. Click Add integration, then pick Slack.
  3. Type an Alias.
  4. Click Cancel.
  5. Click Add integration again and pick Slack.
s7-slack-alias-lost-on-cancel
e2e/specs/webhooks-notifications/s7-slack-alias-lost-on-cancel.spec.ts:8
lines
infisical/frontend/src/pages/organization/SettingsPage/components/OrgWorkflowIntegrationTab/AddWorkflowIntegrationForm.tsx:47
infisical/frontend/src/pages/organization/SettingsPage/components/OrgWorkflowIntegrationTab/SlackIntegrationForm.tsx:150
before the PRs testedno
Opening a project's Applications list counts every certificate on the instance
s10-applications-list-counts-every-certificate-1
pki-enrollmentperformanceloss: time
medium

When You open the Applications list of a Certificate Management project.

Got To show each application's certificate count, the product goes through every certificate assigned to any application on the whole instance, so the list gets slower as other organisations issue certificates.

Want It counts only your project's certificates.

steps
  1. Sign up two organisations; each gets an active Certificate Management project
  2. In the first, create one application and assign no certificates
  3. In the second, create an application and import and assign 5 certificates to it
  4. List the first project's applications (20 per page) and EXPLAIN ANALYZE the product's captured list statement
  5. Assign 45 more certificates in the second organisation and list the first project's applications again
s10-applications-list-counts-every-certificate
e2e/specs/pki-enrollment/s10-applications-list-counts-every-certificate.spec.ts:89
lines
infisical/backend/src/services/pki-application/pki-application-dal.ts:54
infisical/backend/src/services/pki-application/pki-application-dal.ts:39
infisical/backend/src/services/pki-application/pki-application-dal.ts:66
infisical/backend/src/db/migrations/20260504432352_cert-manager-applications.ts:212
before the PRs testedno
A certificate sync can end up filtering on a profile its Application no longer has
s11-sync-filters-on-detached-profile-1
pki-enrollmentbehaviourloss: decision
low

When One admin removes a certificate profile from an Application while another creates a certificate sync that selects certificates by that profile.

Got Both are told it worked. The sync now selects by a profile the Application no longer has.

Want One of the two is told it could not be done.

steps
  1. Sign up a new organization; its Certificate Manager project is active.
  2. Create an internal root CA, a certificate policy, an AWS connection, a certificate profile P and an Application, and attach P to the Application.
  3. Control: on another Application, a sync filtering on an attached profile makes its detach refused; a sync filtering on an unattached profile is refused.
  4. Hold the (Application, P) link row in the database so writes to it wait.
  5. Detach P from the Application; it finds no sync filtering on P and waits on the link delete.
  6. While it waits, create a certificate sync on the Application with the profile filter [P].
  7. Release the hold; count the link rows for (Application, P) and the Application's syncs whose profile filter contains P.
a profile detached while a sync filtering on it is created leaves no sync on a detached profile
e2e/specs/pki-enrollment/s11-sync-filters-on-detached-profile.spec.ts:21
lines
infisical/backend/src/services/pki-application/pki-application-service.ts:560
infisical/backend/src/services/pki-application/pki-application-service.ts:579
infisical/backend/src/services/pki-sync/pki-sync-service.ts:604
infisical/backend/src/services/pki-sync/pki-sync-service.ts:747
before the PRs testedno
An ACME account's order list returns every order it ever placed in one answer
s11-acme-order-list-returns-every-order-1
pki-enrollmentperformanceloss: time
medium

When An ACME client that has renewed many times asks for its list of certificate orders.

Got It receives every order it has ever placed in one answer, and the list only ever grows.

Want One page of recent orders, with a way to fetch the next page.

steps
  1. Create a Certificate Management project with an active root CA, a certificate policy and a certificate profile with ACME enrollment (external account binding skipped).
  2. With an ACME client, read the profile's directory and register a new account.
  3. Place 110 new orders for different domain names from that account.
  4. Ask for the account's order list (POST-as-GET to the orders URL the account answer gave).
s11-acme-order-list-returns-every-order
e2e/specs/pki-enrollment/s11-acme-order-list-returns-every-order.spec.ts:12
lines
infisical/backend/src/ee/services/pki-acme/pki-acme-order-dal.ts:109
infisical/backend/src/ee/services/pki-acme/pki-acme-service.ts:1727
infisical/backend/src/db/migrations/20251104234547_add-pki-acme.ts:82
before the PRs testedno
A SCEP profile's pending-challenge cap is passed when two challenges are requested together
s13-simultaneous-scep-challenges-pass-pending-cap-1
pki-enrollmentbehaviourloss: nothing
low

When Two challenge passwords are requested for a SCEP profile at the same moment.

Got Both are issued, so the profile has more unused challenges than its maximum allows.

Want Only as many challenges as the maximum are issued; the rest are refused.

steps
  1. Create a Certificate Management project with an active root CA and a certificate policy
  2. Create a SCEP certificate profile with dynamic challenges, maximum pending challenges 1
  3. Request a challenge, then another: the second is refused because the cap of 1 is reached (the control)
  4. On a second profile configured the same way, hold inserts into the pending-challenge table and send two challenge requests
  5. Release the hold once both are waiting on their insert
  6. Count the profile's accepted answers and its unexpired pending challenges
a SCEP profile capped at one pending challenge issues one when two are requested together
e2e/specs/pki-enrollment/s13-simultaneous-scep-challenges-pass-pending-cap.spec.ts:18
lines
infisical/backend/src/ee/services/pki-scep/pki-scep-service.ts:1316
infisical/backend/src/ee/services/pki-scep/pki-scep-service.ts:1317
infisical/backend/src/ee/services/pki-scep/pki-scep-service.ts:1324
infisical/backend/src/ee/services/pki-scep/pki-scep-dynamic-challenge-dal.ts:27
before the PRs testedno
ACME clients can't list their orders the way the ACME standard says to
s12-acme-order-list-refuses-post-as-get-1
pki-enrollmentbehaviourloss: time
low

When Your ACME client asks the server for the list of its certificate orders the standard way.

Got The server refuses the request as malformed, so the client cannot see its orders.

Want The server returns the account's orders.

steps
  1. Create a Certificate Management project with an active root CA, a certificate policy and an ACME-enabled certificate profile that skips external account binding
  2. With an ACME client, read the profile's directory, register a new account and place one order
  3. Fetch the account's orders URL with a POST-as-GET request (a signed JWS whose payload is the empty string)
lines
infisical/backend/src/ee/services/pki-acme/pki-acme-service.ts:323
infisical/backend/src/ee/services/pki-acme/pki-acme-schemas.ts:133
infisical/backend/src/ee/routes/v1/pki-acme-router.ts:357
before the PRs testedno
A certificate sync's details page says the sync does not exist when loading it fails
s1-failed-certificate-sync-shown-as-not-found-1
uiloss: time
medium

When

Got

Want

steps
  1. Sign in as a member of a Certificate Manager project that is the organization's active project
  2. Create an application and an AWS Certificate Manager sync in it
  3. Open the sync's details page (Integrations > the sync) and see it load
  4. Make the request that loads the sync's details fail with a server error, and reload the page
  5. Look at the page once loading ends
lines
infisical/frontend/src/pages/cert-manager/PkiSyncDetailsByIDPage/PkiSyncDetailsByIDPage.tsx:59
infisical/frontend/src/pages/cert-manager/PkiSyncDetailsByIDPage/PkiSyncDetailsByIDPage.tsx:70
before the PRs testedno
A certificate sync whose certificate list fails to load says no certificates are part of the sync
s1-failed-sync-certificates-shown-as-none-1
uiloss: decision
medium

When

Got

Want

steps
  1. Create a Certificate Manager project, an application in it, import one certificate into the application, an AWS app connection, and an AWS Certificate Manager sync holding that certificate
  2. Make the sync's certificate list request fail with a server error (both the first try and the retry)
  3. Open the sync's details page
  4. Look at the Certificates section
lines
infisical/frontend/src/pages/cert-manager/PkiSyncDetailsByIDPage/components/PkiSyncCertificatesSection.tsx:107
infisical/frontend/src/pages/cert-manager/PkiSyncDetailsByIDPage/components/PkiSyncCertificatesSection.tsx:233
before the PRs testedno
The discovered installations list slows as discovery finds more installations anywhere
s14-installations-list-unindexed-lookups-1
pki-syncs-alertsperformanceloss: time
medium

When You open the discovered certificate installations list, or filter it by a certificate.

Got The product looks through the discovery records of every project on the instance to label each installation, so the list gets slower as discovery runs anywhere.

Want Each installation's discovery job and certificates are found directly, at the same cost however much discovery has run.

steps
  1. In a Certificate Manager project, open Certificate Discovery and its Installations tab (the discovered installations list)
  2. Optionally filter the list by one certificate
  3. Look at the indexes on the discovery-installation link table (pki_discovery_installations) and the installation-certificate link table (pki_certificate_installation_certs) in the database the product's migrations built
s14 installations list finds each installation's discovery job through an index
e2e/specs/pki-syncs-alerts/s14-installations-list-unindexed-lookups.spec.ts:7
lines
infisical/backend/src/ee/services/pki-discovery/pki-certificate-installation-dal.ts:73
infisical/backend/src/ee/services/pki-discovery/pki-certificate-installation-dal.ts:100
infisical/backend/src/ee/services/pki-discovery/pki-certificate-installation-dal.ts:200
infisical/backend/src/db/migrations/20260205012343_add-pki-discovery-tables.ts:96
infisical/backend/src/db/migrations/20260205012343_add-pki-discovery-tables.ts:112
before the PRs testedno
A failed alert list load is shown as having no alerts configured
s14-failed-alert-list-shown-as-no-alerts-1
pki-syncs-alertsuiloss: decision
medium

When You open the Alerting screen and the alerts cannot be loaded.

Got The screen says no PKI alerts are configured yet, with no sign anything went wrong.

Want The screen tells you the alerts could not be loaded.

steps
  1. Sign in and open a Certificate Manager project's Alerting screen; the alert list loads.
  2. Make the alert list request fail with a server error (the query retries once, so both attempts fail).
  3. Reload the Alerting screen and wait for the list to settle.
s14-failed-alert-list-shown-as-no-alerts
e2e/specs/pki-syncs-alerts/s14-failed-alert-list-shown-as-no-alerts.spec.ts:10
lines
infisical/frontend/src/views/PkiAlertsV2Page/PkiAlertsV2Page.tsx:46
infisical/frontend/src/views/PkiAlertsV2Page/PkiAlertsV2Page.tsx:120
before the PRs testedno
With more than 25 legacy certificate alerts, the Alerting screen's pager shows the page count as the alert count and cannot reach page 2
s1-alerts-past-first-page-unreachable-1
uiloss: time
medium

When

Got

Want

steps
  1. In the organization's active Certificate Manager project, have more than 25 legacy certificate alerts (26 or more, made before legacy alert writes were blocked)
  2. Open the project's Alerting screen (25 alerts per page)
  3. Look at the pager under the Certificate Alerts table
lines
infisical/frontend/src/views/PkiAlertsV2Page/PkiAlertsV2Page.tsx:72
infisical/frontend/src/views/PkiAlertsV2Page/PkiAlertsV2Page.tsx:176
infisical/frontend/src/components/v2/Pagination/Pagination.tsx:42
before the PRs testedno
A certificate's sync dialog gets slower with every sync in the project
s14-certificate-sync-dialog-reads-per-sync-1
pki-syncs-alertsperformanceloss: time
medium

When You open a certificate's Manage Syncs dialog in a project with many certificate syncs.

Got The dialog does one database lookup per sync, so it loads slower as syncs are added.

Want The dialog costs the same whether the project has 3 syncs or 30.

steps
  1. In the org's active Certificate Manager project, create an application with 3 AWS Certificate Manager syncs and another with 30
  2. List each application's syncs for one certificate (the read the certificate's Manage Syncs dialog makes, with a certificate id)
  3. Count the database statements each list makes
s14-certificate-sync-dialog-reads-per-sync
e2e/specs/pki-syncs-alerts/s14-certificate-sync-dialog-reads-per-sync.spec.ts:16
lines
infisical/backend/src/services/pki-sync/pki-sync-service.ts:1184
infisical/backend/src/services/pki-sync/pki-sync-service.ts:1187
infisical/backend/src/services/certificate-sync/certificate-sync-dal.ts:17
infisical/frontend/src/pages/cert-manager/CertificatesPage/components/CertificateManagePkiSyncsModal.tsx:76
before the PRs testedno
The certificate Integrations tab downloads every sync of a project to show 20
s15-certificate-sync-list-loads-every-sync-1
pki-syncs-alertsperformanceloss: time
medium

When Your Certificate Manager project has many certificate syncs and you open its Integrations tab.

Got The page downloads every sync in the project, then shows 20. It downloads them all again every 30 seconds.

Want The page loads only the 20 syncs it shows.

steps
  1. Sign in to an org whose active Certificate Manager project has an application with more than 20 certificate syncs (25 made in the repro; the project held 174)
  2. Open the project's Integrations page on the certificate syncs tab; the table shows 20 per page
  3. Watch the sync list the page downloads, and again every 30 seconds while the tab stays open
the certificate syncs list answers at most one page of syncs
e2e/specs/pki-syncs-alerts/s15-certificate-sync-list-loads-every-sync.spec.ts:9
lines
infisical/backend/src/server/routes/v1/pki-sync-routers/pki-sync-router.ts:160
infisical/backend/src/services/pki-sync/pki-sync-service.ts:1173
infisical/backend/src/services/pki-sync/pki-sync-dal.ts:207
infisical/frontend/src/pages/cert-manager/IntegrationsListPage/components/PkiSyncsTab/PkiSyncsTab.tsx:85
infisical/frontend/src/pages/cert-manager/IntegrationsListPage/components/PkiSyncsTab/PkiSyncTable/PkiSyncsTable.tsx:370
before the PRs testedno
A failed load of the Certificate Discovery jobs or installations shows "No discovery jobs defined" / "No installations found" instead of an error
s1-failed-discovery-lists-shown-as-empty-1
behaviourloss: decision
medium

When

Got

Want

steps
  1. Sign in as a member of a Certificate Manager project
  2. Have the discovery job list fail to load on the server (500)
  3. Open the project's Certificate Discovery page: the Jobs tab says "No discovery jobs defined" with no error
  4. Have the installation list fail to load on the server (500) and open the Installations tab: it says "No installations found" with no error
lines
infisical/frontend/src/pages/cert-manager/DiscoveryPage/components/DiscoveryJobsTab.tsx:95
infisical/frontend/src/pages/cert-manager/DiscoveryPage/components/DiscoveryJobsTab.tsx:217
infisical/frontend/src/pages/cert-manager/DiscoveryPage/components/InstallationsTab.tsx:69
infisical/frontend/src/pages/cert-manager/DiscoveryPage/components/InstallationsTab.tsx:120
before the PRs testedno
Saving a certificate sync's edit form removes a certificate added since it was opened
s16-stale-sync-edit-unlinks-added-certificate-1
pki-syncs-alertsbehaviourloss: work
low

When You edit a certificate sync while someone else adds a certificate to it.

Got Your save succeeds, and their certificate is quietly removed from the sync.

Want Their certificate stays in the sync, or you are told the sync changed.

steps
  1. Create an AWS Certificate Manager sync holding certificate A, with certificates B and C in the same application
  2. Add B to the sync, open the sync's edit form and save it: A and B stay linked (control)
  3. Open the sync's edit form (it loads the sync's certificate filters)
  4. In another tab, add certificate C to the sync
  5. Save the open edit form with a new description
  6. List the sync's certificates
s16-stale-sync-edit-unlinks-added-certificate
e2e/specs/pki-syncs-alerts/s16-stale-sync-edit-unlinks-added-certificate.spec.ts:20
lines
infisical/frontend/src/components/pki-syncs/forms/EditPkiSyncForm.tsx:234
infisical/frontend/src/components/pki-syncs/forms/EditPkiSyncForm.tsx:314
infisical/backend/src/services/pki-sync/pki-sync-service.ts:1046
infisical/backend/src/services/pki-sync/pki-sync-service.ts:1070
before the PRs testedno
A retransmitted SCEP enrollment request creates a duplicate certificate request
s14-scep-retransmission-duplicate-request-1
pki-enrollmentbehaviourloss: time
medium

When A device's SCEP enrollment answer is lost and the device resends the same request

Got The system opens a second approval request and a second certificate request for the same device

Want The resent request is recognised and leaves one request to approve

steps
  1. Create a Certificate Management project with an active internal CA
  2. Add a SCEP certificate profile with a static challenge password
  3. Add a certificate-request approval policy that matches the profile so enrollment needs approval
  4. As a SCEP device, send a PKCSReq enrollment request and let its answer be lost
  5. Send the identical PKCSReq again with the same transaction id (the retransmission)
  6. Count the pending certificate requests and approval requests for the device: there are two of each, not one
a retransmitted SCEP PKCSReq leaves one certificate request, not two
e2e/specs/pki-enrollment/s14-scep-retransmission-duplicate-request.spec.ts:82
lines
infisical/backend/src/ee/services/pki-scep/pki-scep-service.ts:887
infisical/backend/src/ee/services/pki-scep/pki-scep-service.ts:900
infisical/backend/src/ee/services/pki-scep/pki-scep-service.ts:901
infisical/backend/src/services/certificate-v3/certificate-v3-service.ts:1502
before the PRs testedno
A load-balancer sync can end up with two default certificates
s16-load-balancer-sync-two-default-certificates-1
pki-syncs-alertsbehaviourloss: time
low

When Two people choose different default certificates for one load-balancer sync at the same moment.

Got Both choices are kept, and the sync shows two default certificates.

Want Only one certificate is the default — the later choice.

steps
  1. In a Certificate Manager application, create an AWS Elastic Load Balancer sync and add three certificates A, B and C to it
  2. Make C the sync's default certificate
  3. From two tabs (or two admins), choose A and B as the sync's default at the same moment
  4. Open the sync's certificate list
s16-load-balancer-sync-two-default-certificates
e2e/specs/pki-syncs-alerts/s16-load-balancer-sync-two-default-certificates.spec.ts:37
lines
infisical/backend/src/services/pki-sync/pki-sync-service.ts:1936
infisical/backend/src/services/pki-sync/pki-sync-service.ts:1941
infisical/backend/src/services/certificate-sync/certificate-sync-dal.ts:242
infisical/backend/src/services/certificate-sync/certificate-sync-dal.ts:212
before the PRs testedno
Two discovery jobs in one project can both start scanning at once
s16-discovery-scans-start-together-1
pki-syncs-alertsbehaviourloss: nothing
low

When You start the scans of two discovery jobs in the same project at the same moment.

Got Both scans start, though the product says only one scan per project can run at a time.

Want One scan starts and the other is refused with the one-scan-per-project message.

steps
  1. Create a Certificate Manager project with two discovery jobs D1 and D2 targeting localhost:22380
  2. Control: trigger D1's scan, then D2's scan — D2 is refused with "Another scan is already running in this project. Only one concurrent scan per project is allowed."
  3. In a second fresh project with jobs D1 and D2, hold D1's row in pki_discovery_configs (SELECT ... FOR UPDATE) so D1's scan claim parks
  4. Trigger D1's scan (it waits on the hold), then trigger D2's scan while held, then release the hold
s16 two discovery jobs of one project triggered together do not both start a scan
e2e/specs/pki-syncs-alerts/s16-discovery-scans-start-together.spec.ts:24
lines
infisical/backend/src/ee/services/pki-discovery/pki-discovery-config-dal.ts:125
infisical/backend/src/ee/services/pki-discovery/pki-discovery-config-dal.ts:129
infisical/backend/src/ee/services/pki-discovery/pki-discovery-service.ts:446
infisical/backend/src/ee/services/pki-discovery/pki-discovery-service.ts:449
before the PRs testedno
A failed load of an HSM connector's linked resources tells the admin the connector is safe to delete
s3-hsm-connector-linked-resources-failure-shown-as-safe-to-delete-1
uiloss: decision
medium

When

Got

Want

steps
  1. Sign in as a project admin of a certificate-manager project
  2. Open an HSM connector's page
  3. Have the linked-resources load fail (a server error, seen after the one automatic retry)
  4. Look at the Linked resources card
lines
infisical/frontend/src/pages/cert-manager/HsmConnectorDetailsByIDPage/components/HsmConnectorLinkedResourcesSection.tsx:38
infisical/frontend/src/pages/cert-manager/HsmConnectorDetailsByIDPage/components/HsmConnectorLinkedResourcesSection.tsx:75
infisical/frontend/src/hooks/api/hsmConnectors/queries.tsx:52
infisical/backend/src/services/hsm-connector/hsm-connector-service.ts:377
before the PRs testedno
A failed load of the code signers list shows "No signers yet" instead of an error
s1-signers-load-failure-shown-as-no-signers-1
uiloss: decision
medium

When

Got

Want

steps
  1. Sign in as a member of a Certificate Manager project
  2. Open Code Signing in that project
  3. Have the signers list fail to load (a server error, seen after the one automatic retry)
  4. Look at the Signers card
lines
infisical/frontend/src/pages/cert-manager/CodeSigningPage/components/SignersTable.tsx:79
infisical/frontend/src/pages/cert-manager/CodeSigningPage/components/SignersTable.tsx:211
infisical/frontend/src/hooks/api/signers/queries.tsx:66
before the PRs testedno
A failed signing history load tells the user a signer has "No signing activity yet"
s2-signing-history-failure-shown-as-no-activity-1
uiloss: decision
medium

When

Got

Want

steps
  1. Sign in as a project admin of a certificate-manager project
  2. Create a code signer on an imported code-signing certificate and sign one digest with it
  3. Have the signing history load fail (a server error, seen after the one automatic retry)
  4. Open the signer on its Activity tab
lines
infisical/frontend/src/pages/cert-manager/SignerDetailPage/components/SigningOperationsTable.tsx:70
infisical/frontend/src/pages/cert-manager/SignerDetailPage/components/SigningOperationsTable.tsx:218
infisical/frontend/src/hooks/api/signers/queries.tsx:116
before the PRs testedno
Edit KMS Credentials opens empty, with no form and no error
s16-kms-credentials-load-failure-shown-blank-1
kmsuiloss: time
medium

When You open Edit Credentials on an external KMS and it cannot be loaded.

Got The dialog opens with only its title and description. There is no form and no message explaining why.

Want The dialog tells you the KMS could not be loaded.

steps
  1. Sign in as an org admin of an organisation with an AWS external KMS
  2. Open Organisation Settings, Encryption tab
  3. Have the server fail the read of that external KMS
  4. Open the KMS row's Actions menu and choose Edit Credentials
  5. Wait for the load and its retry to fail
a failed external KMS load shows an error in Edit KMS Credentials
e2e/specs/kms/s16-kms-credentials-load-failure-shown-blank.spec.ts:9
lines
infisical/frontend/src/pages/organization/SettingsPage/components/OrgEncryptionTab/EditExternalKmsCredentialsModal.tsx:28
infisical/frontend/src/pages/organization/SettingsPage/components/OrgEncryptionTab/EditExternalKmsCredentialsModal.tsx:46
before the PRs testedno
The alerts list does more work the more often each alert has run
s17-alerts-list-reads-every-alert-run-1
pki-syncs-alertsperformanceloss: time
low

When You open the Alerting screen of a project whose legacy alerts have run many times.

Got To show each alert's last run, the list goes through every past run of every alert on the page, so it does more work as the alerts keep running.

Want The list looks up only each alert's newest run.

steps
  1. In the org's active Certificate Manager project, have a legacy revocation alert in an application, with an email channel
  2. Issue a certificate, assign it to the application and revoke it, 5 times: the alert gets 5 past runs
  3. List the project's alerts filtered to that alert, and EXPLAIN (ANALYZE) the list's last-run statement with its alert id
  4. Revoke 55 more certificates in the application (60 runs) and list again
s17 the alerts list reads one past run per alert, not its whole run history
e2e/specs/pki-syncs-alerts/s17-alerts-list-reads-every-alert-run.spec.ts:14
lines
infisical/backend/src/services/pki-alert-v2/pki-alert-v2-dal.ts:256
infisical/backend/src/services/pki-alert-v2/pki-alert-v2-dal.ts:265
infisical/backend/src/services/pki-alert-v2/pki-alert-v2-service.ts:861
before the PRs testedno
Saving a signer's approval policy twice at once keeps both versions' steps
s10-overlapping-policy-saves-keep-both-steps-1
code-signingbehaviourloss: work
low

When Two admins save the same signer's approval policy at the same moment.

Got Both saves succeed and the policy now lists two 'Step 1's, one from each save.

Want The policy holds only the steps of the last save.

steps
  1. Create a code signer in a Certificate Manager project; its approval policy is created empty
  2. Save a one-step policy (the signer's admin as approver, max 5 signings) twice in turn: the policy has one step
  3. Hold the policy's existing step row from another database session (select ... for update), then send two one-step saves named 'Save A' and 'Save B' for the same signer
  4. Release the hold once both saves are waiting on it, and read the signer's approval policy
two overlapping approval-policy saves leave the signer with one step
e2e/specs/code-signing/s10-overlapping-policy-saves-keep-both-steps.spec.ts:14
lines
infisical/backend/src/services/signer/signer-policy-service.ts:493
infisical/backend/src/services/signer/signer-policy-service.ts:494
infisical/backend/src/services/signer/signer-policy-service.ts:497
infisical/backend/src/services/signer/signer-policy-service.ts:502
infisical/backend/src/db/migrations/20251203002657_global-approvals.ts:39
before the PRs testedno
The same request to sign sent twice at once opens two pending requests
s10-concurrent-requests-to-sign-open-duplicate-pending-1
code-signingbehaviourloss: time
low

When The same request to sign reaches the signer twice at the same moment, from two tabs or a build tool retrying.

Got Two separate pending requests appear for the approvers, though sending it again later returns the one already open.

Want Only one pending request is opened, and the second send returns it.

steps
  1. Create a code signer in a Certificate Manager project and give it a one-step approval policy with yourself as approver
  2. Send a request to sign with a justification, 2 signings and a hostname scope, then send the identical request again: the second returns the same pending request
  3. On a second such signer, send the identical request to sign twice at the same moment (forced by holding the signer's approval policy row so both duplicate lookups run before either insert)
  4. Count the pending approval requests for that signer
identical requests to sign sent together open one pending request
e2e/specs/code-signing/s10-concurrent-requests-to-sign-open-duplicate-pending.spec.ts:27
lines
infisical/backend/src/services/signer/signer-policy-service.ts:268
infisical/backend/src/services/signer/signer-policy-service.ts:630
infisical/backend/src/services/signer/signer-policy-service.ts:639
infisical/backend/src/services/signer/signer-policy-service.ts:650
infisical/backend/src/services/approval-policy/approval-request-fns.ts:131
before the PRs testedno
A signer with members looks like it has none when its members fail to load
s10-signer-members-failure-shown-as-no-members-1
code-signinguiloss: decision
medium

When You open a signer's Members tab and its members can't be loaded.

Got It says 'No members yet', as if nobody had access to the signer.

Want It tells you the members couldn't be loaded.

steps
  1. Create a code signer in a Certificate Manager project and add a user as a member.
  2. Make the signer's member-list requests (users, machine identities, groups) fail with a server error.
  3. Open the signer and go to its Members tab.
  4. Wait past the one automatic retry.
s10-signer-members-failure-shown-as-no-members
e2e/specs/code-signing/s10-signer-members-failure-shown-as-no-members.spec.ts:6
lines
infisical/frontend/src/hooks/api/signers/queries.tsx:153
infisical/frontend/src/pages/cert-manager/SignerDetailPage/components/SignerMembersTab.tsx:119
infisical/frontend/src/pages/cert-manager/SignerDetailPage/components/SignerMembersTab.tsx:172
before the PRs testedno
A signer's request list gets slower with every approved signing in the instance
s11-signer-requests-list-reads-every-approved-signing-1
code-signingperformanceloss: time
medium

When You open the requests of a code signer that has at least one approved request.

Got Loading them does work for every approved signing made by every signer in the whole instance, so it keeps getting slower as anyone signs.

Want Loading a signer's requests only looks at that signer's own signings.

steps
  1. In a Certificate Manager project, create signer A and signer B, each with a one-step approval policy.
  2. Pre-approve the admin on signer A (no signings made) and on signer B for 60 signings.
  3. Sign 5 times with signer B, then load signer A's requests to sign and replay the list's own statement under EXPLAIN ANALYZE.
  4. Sign 45 more times with signer B and load signer A's requests again.
s11-signer-requests-list-reads-every-approved-signing
e2e/specs/code-signing/s11-signer-requests-list-reads-every-approved-signing.spec.ts:47
lines
infisical/backend/src/services/signer/signer-request-dal.ts:30
infisical/backend/src/services/signer/signer-request-dal.ts:44
infisical/backend/src/services/signer/signing-operation-dal.ts:133
before the PRs testedno
A revoked signing request can still be used to sign
s12-signing-request-revoked-during-approval-still-signs-1
code-signingbehaviourloss: decision
medium

When An admin revokes a signing request at the same moment another approver approves it.

Got The request shows as cancelled, but the requester keeps a working approval and can still sign.

Want A revoked request leaves no approval behind and signing is refused.

steps
  1. Create a code signer with a one-step approval policy whose approver is the admin, and add a machine identity as an operator.
  2. As the machine identity, open a request to sign (1 signing).
  3. Hold the request's row in the database so the admin's approve waits on it.
  4. As the admin, approve the request; while it waits, revoke the same request.
  5. Release the hold, then read the request and its approvals, and sign as the machine identity.
s12-signing-request-revoked-during-approval-still-signs
e2e/specs/code-signing/s12-signing-request-revoked-during-approval-still-signs.spec.ts:33
lines
infisical/backend/src/services/signer/signer-policy-service.ts:821
infisical/backend/src/services/signer/signer-policy-service.ts:827
infisical/backend/src/services/signer/signer-policy-service.ts:843
infisical/backend/src/services/approval-policy/approval-policy-service.ts:1358
infisical/backend/src/services/approval-policy/code-signing/code-signing-policy-factory.ts:169
infisical/backend/src/services/signer/signer-service.ts:1479
before the PRs testedno
Adding users to a code signer gets slower with every user added at once
s12-adding-signer-users-rereads-per-user-1
code-signingperformanceloss: time
low

When An admin adds several users to a code signer at once from the Add member dialog.

Got The product checks the signer and each user's memberships again for every user, so the add takes longer the more users are picked.

Want Adding 12 users costs about the same checks as adding 2, plus the new memberships.

steps
  1. As a Certificate Manager project admin, invite 12 users to the organisation and add them to the project as members.
  2. Create a code signer and add 2 of those users to it in one add, counting the database statements; repeat on fresh signers three times.
  3. Create another signer and add all 12 users in one add, counting the statements; repeat on fresh signers three times.
  4. Compare the signer and membership lookups of the two sizes.
s12-adding-signer-users-rereads-per-user
e2e/specs/code-signing/s12-adding-signer-users-rereads-per-user.spec.ts:11
lines
infisical/backend/src/services/signer-membership/signer-membership-service.ts:722
infisical/backend/src/services/signer-membership/signer-membership-service.ts:727
infisical/backend/src/services/signer-membership/signer-membership-service.ts:179
infisical/backend/src/services/signer-membership/signer-membership-service.ts:181
infisical/backend/src/services/signer-membership/signer-membership-service.ts:197
infisical/backend/src/services/signer-membership/signer-membership-service.ts:209
infisical/backend/src/services/signer-membership/signer-membership-service.ts:225
before the PRs testedno
A member removed while a signer's policy saves stays an approver
s11-member-removed-mid-save-stays-approver-1
code-signingbehaviourloss: work
medium

When You remove someone from a signer at the same moment another admin saves the signer's approval policy with that person as an approver.

Got They are removed as a member, yet they stay listed as an approver on the policy.

Want They are either kept off the policy, or dropped as an approver when removed.

steps
  1. In a Certificate Manager project create a code signer and add the admin and a second user as its members
  2. Give the signer an approval policy of one step whose approver is the admin
  3. Confirm the guards in turn: with the second user removed from the signer, a save naming them as approver is refused; with them a member and an approver, removing them clears their approver row
  4. Hold the policy's approver rows from another database session so the next save parks at its delete of the old approver rows
  5. Save the policy naming both the admin and the second user as approvers, and while it is parked remove the second user from the signer (answers success)
  6. Release the hold and read the policy's approver rows and the second user's signer membership
a member removed while its approval policy saves does not stay an approver
e2e/specs/code-signing/s11-member-removed-mid-save-stays-approver.spec.ts:12
lines
infisical/backend/src/services/signer/signer-policy-service.ts:464
infisical/backend/src/services/signer/signer-policy-service.ts:493
infisical/backend/src/services/signer/signer-policy-service.ts:513
infisical/backend/src/services/signer-membership/signer-membership-service.ts:613
before the PRs testedno
A failed scanning configuration load says "Could not find Project Configuration" instead of reporting the error
s4-failed-scanning-config-shown-as-not-found-1
behaviourloss: time
medium

When

Got

Want

steps
  1. Sign in as a member of a Secret Scanning project
  2. Make the scanning configuration request fail with a server error (both the request and its retry)
  3. Open the project Settings on the Scanning Settings tab
  4. Read the tab once the requests have failed
lines
infisical/frontend/src/pages/secret-scanning/SettingsPage/components/ProjectScanningConfigTab/ProjectScanningConfigTab.tsx:13
infisical/frontend/src/pages/secret-scanning/SettingsPage/components/ProjectScanningConfigTab/ProjectScanningConfigTab.tsx:36
infisical/backend/src/ee/services/secret-scanning-v2/secret-scanning-v2-service.ts:877
before the PRs testedno
A secret scanning data source that fails to load is reported as not existing
s3-failed-data-source-shown-as-not-found-1
behaviourloss: decision
medium

When

Got

Want

steps
  1. Sign in as a member of a Secret Scanning project that has a GitHub data source
  2. Open the data source's page (Data Sources, then the data source) while its details request fails with a server error, including the app's retry
  3. Read what the page says
lines
infisical/frontend/src/pages/secret-scanning/SecretScanningDataSourceByIdPage/SecretScanningDataSourceByIdPage.tsx:34
infisical/frontend/src/pages/secret-scanning/SecretScanningDataSourceByIdPage/SecretScanningDataSourceByIdPage.tsx:46
before the PRs testedno
A failed Secret Scanning findings load tells the user the project has not uncovered any secret leaks
s1-failed-findings-shown-as-no-leaks-1
behaviourloss: decision
high

When

Got

Want

steps
  1. Sign in as a member of a Secret Scanning project that has findings
  2. Make the findings list request fail with a server error (on both the first try and the client's retry)
  3. Open the project's Findings page
  4. The page shows the Findings table with "This project has not uncovered any secret leaks" and no error
lines
infisical/frontend/src/pages/secret-scanning/SecretScanningFindingsPage/components/SecretScanningFindingsSection.tsx:11
infisical/frontend/src/pages/secret-scanning/SecretScanningFindingsPage/components/SecretScanningFindingsTable.tsx:415
infisical/frontend/src/hooks/api/secretScanningV2/queries.tsx:254
before the PRs testedno
Adding several machine identities to a signer sends one add per identity
s13-adding-signer-identities-one-request-each-1
code-signingperformanceloss: time
low

When You add several machine identities or groups to a signer at once.

Got The page adds them one at a time, one call to the server for each.

Want All the selected members are added in one go, as users are.

steps
  1. Create a code signer in a Certificate Manager project that has several machine identities as members
  2. Open the signer's Members tab and click Add Member
  3. Choose Machine Identities and select 6 identities
  4. Click Add members and watch the requests the page sends
s13 adding several machine identities to a signer sends one request per submit
e2e/specs/code-signing/s13-adding-signer-identities-one-request-each.spec.ts:8
lines
infisical/frontend/src/pages/cert-manager/SignerDetailPage/components/AddSignerMemberModal.tsx:148
infisical/frontend/src/pages/cert-manager/SignerDetailPage/components/AddSignerMemberModal.tsx:156
infisical/frontend/src/pages/cert-manager/SignerDetailPage/components/AddSignerMemberModal.tsx:140
infisical/backend/src/server/routes/v1/signer-membership-routers/identities-router.ts:59
before the PRs testedno
A failed data sources load tells the user the Secret Scanning project has no data sources configured
s1-data-sources-failure-shown-as-empty-1
behaviourloss: decision
medium

When

Got

Want

steps
  1. Sign in as a member of an organisation with a Secret Scanning project that has a GitHub data source and findings
  2. Have the data sources list request fail with a server error (and its one automatic retry)
  3. Open the project's Data Sources page
  4. The table says "This project has no data sources configured" while the sidebar shows 3 findings, and no error is shown
lines
infisical/frontend/src/pages/secret-scanning/SecretScanningDataSourcesPage/components/SecretScanningDataSourcesSection.tsx:26
infisical/frontend/src/pages/secret-scanning/SecretScanningDataSourcesPage/components/SecretScanningDataSourcesTable.tsx:345
infisical/frontend/src/hooks/api/secretScanningV2/queries.tsx:124
before the PRs testedno
A data source's repositories or scans that fail to load are shown as none
s2-data-source-scans-and-repositories-failure-shown-as-empty-1
uiloss: decision
medium

When

Got

Want

steps
  1. Sign in as a member of a Secret Scanning project that has a GitHub data source with at least one repository and one scan
  2. Make the data source's repositories request (or, separately, its scan history request) fail with a server error on every attempt
  3. Open the data source's page from Data Sources
  4. Wait for the request and its one retry to fail, then look at the Repositories table (or the Scan History)
lines
infisical/frontend/src/pages/secret-scanning/SecretScanningDataSourceByIdPage/components/SecretScanningResourceTable.tsx:81
infisical/frontend/src/pages/secret-scanning/SecretScanningDataSourceByIdPage/components/SecretScanningResourceTable.tsx:341
infisical/frontend/src/pages/secret-scanning/SecretScanningDataSourceByIdPage/components/SecretScanningScanTable.tsx:69
infisical/frontend/src/pages/secret-scanning/SecretScanningDataSourceByIdPage/components/SecretScanningScanTable.tsx:286
before the PRs testedno
A refused bulk status update still changes the other selected findings
s10-refused-bulk-finding-update-still-applied-1
secret-scanningbehaviourloss: decision
medium

When You mark several findings at once and one of them was removed in the meantime.

Got You are told the update failed, but the other findings were changed anyway.

Want Either none of the findings change, or you are told which ones did.

steps
  1. As a member of a Secret Scanning project, have two unresolved findings on one data source and a third finding on another data source
  2. Delete the second data source (its finding goes with it)
  3. Send one bulk status update marking all three findings as false positive
  4. The update is refused: "Could not find Secret Scanning Finding with ID ..."
  5. Read the two remaining findings' status
a bulk status update refused for a removed finding changes none of the other selected findings
e2e/specs/secret-scanning/s10-refused-bulk-finding-update-still-applied.spec.ts:59
lines
infisical/backend/src/ee/routes/v2/secret-scanning-v2-routers/secret-scanning-v2-router.ts:311
infisical/backend/src/ee/routes/v2/secret-scanning-v2-routers/secret-scanning-v2-router.ts:329
infisical/backend/src/ee/services/secret-scanning-v2/secret-scanning-v2-service.ts:829
infisical/frontend/src/hooks/api/secretScanningV2/mutations.tsx:155
before the PRs testedno
The Findings page downloads every finding the project ever had to show 20
s10-findings-list-loads-every-finding-1
secret-scanningperformanceloss: time
medium

When Your Secret Scanning project has collected many findings over time.

Got The Findings page downloads all of them, with full details, each time it opens and every 30 seconds, to show 20.

Want The page loads only the findings it shows.

steps
  1. Have a Secret Scanning project whose repositories have produced a few hundred findings
  2. Open the project's Findings page
  3. Watch the findings list request the page makes on load and again every 30 seconds
the Findings list answers one page of findings, not every finding of the project
e2e/specs/secret-scanning/s10-findings-list-loads-every-finding.spec.ts:63
lines
infisical/backend/src/ee/services/secret-scanning-v2/secret-scanning-v2-dal.ts:533
infisical/backend/src/ee/services/secret-scanning-v2/secret-scanning-v2-service.ts:788
infisical/backend/src/ee/routes/v2/secret-scanning-v2-routers/secret-scanning-v2-router.ts:183
infisical/backend/src/ee/routes/v2/secret-scanning-v2-routers/secret-scanning-v2-router.ts:212
infisical/frontend/src/pages/secret-scanning/SecretScanningFindingsPage/components/SecretScanningFindingsTable.tsx:173
infisical/frontend/src/pages/secret-scanning/SecretScanningFindingsPage/components/SecretScanningFindingsSection.tsx:14
before the PRs testedno
A data source's Scan History downloads every scan it ever ran to show 10
s10-scan-history-loads-every-scan-1
secret-scanningperformanceloss: time
medium

When You open a GitHub data source that has been scanned many times.

Got Its Scan History downloads every scan ever run, and again every 30 seconds, though only 10 are shown.

Want Only the page of scans you are looking at is loaded.

steps
  1. Have a GitHub data source in a Secret Scanning project whose repository has received 200 pushes (200 realtime scans, two findings each)
  2. Open the data source's page so its Scans table loads the scan history
  3. Watch the scan history request the page makes (and repeats every 30 seconds)
a data source's scan history answers one page of scans
e2e/specs/secret-scanning/s10-scan-history-loads-every-scan.spec.ts:52
lines
infisical/backend/src/ee/services/secret-scanning-v2/secret-scanning-v2-dal.ts:393
infisical/backend/src/ee/services/secret-scanning-v2/secret-scanning-v2-service.ts:732
infisical/backend/src/ee/routes/v2/secret-scanning-v2-routers/secret-scanning-v2-endpoints.ts:579
infisical/frontend/src/pages/secret-scanning/SecretScanningDataSourceByIdPage/components/SecretScanningScanTable.tsx:269
before the PRs testedno
Changing a finding's status erases remarks another member saved meanwhile
s11-status-only-save-erases-newer-remarks-1
secret-scanningbehaviourloss: work
medium

When You open Update Status on a finding and a teammate adds remarks to it before you save.

Got You change only the status and save. The teammate's remarks are replaced by the old ones, with no warning.

Want Your status change is saved and your teammate's remarks are kept.

steps
  1. Open a Secret Scanning project's Findings page, expand an unresolved finding whose remarks read 'r0' and open Update Status on it
  2. While the dialog is open, another member saves the remarks 'r1' on the same finding
  3. In the open dialog, change only the status to False Positive and save
  4. Read the finding's remarks
a status-only save from the Update Status dialog keeps remarks another member saved after it opened
e2e/specs/secret-scanning/s11-status-only-save-erases-newer-remarks.spec.ts:57
lines
infisical/frontend/src/pages/secret-scanning/SecretScanningFindingsPage/components/SecretScanningUpdateFindingModal.tsx:51
infisical/frontend/src/pages/secret-scanning/SecretScanningFindingsPage/components/SecretScanningUpdateFindingModal.tsx:68
infisical/backend/src/ee/services/secret-scanning-v2/secret-scanning-v2-service.ts:830
before the PRs testedno
A finding reopened while being resolved stays unresolved with a resolved date
s11-overlapping-status-changes-leave-resolved-date-1
secret-scanningbehaviourloss: nothing
low

When Two people change the same finding's status at the same moment, one to Resolved and one to Unresolved.

Got The finding shows Unresolved but still records a resolved date and the first person as its triager.

Want An unresolved finding carries no resolved date, whatever order the two changes land in.

steps
  1. Have an unresolved Secret Scanning finding in a project.
  2. As one member, change its status to Resolved.
  3. At the same moment, as another member, change its status to Unresolved, so both changes read the finding before either saves.
  4. Read the finding back through the API.
a finding changed to resolved and back to unresolved at the same moment keeps no resolved date
e2e/specs/secret-scanning/s11-overlapping-status-changes-leave-resolved-date.spec.ts:55
lines
infisical/backend/src/ee/services/secret-scanning-v2/secret-scanning-v2-service.ts:805
infisical/backend/src/ee/services/secret-scanning-v2/secret-scanning-v2-service.ts:826
infisical/backend/src/ee/services/secret-scanning-v2/secret-scanning-v2-service.ts:838
before the PRs testedno
The Data Sources and Repositories lists read every scan ever run to show one row each
s13-dashboards-read-every-scan-1
secret-scanningperformanceloss: time
medium

When Your repositories have been scanned many times, as every push triggers a scan.

Got The Data Sources and Repositories lists fetch every scan ever run, every 30 seconds, to show one line each.

Want Each list costs the same however long the scan history is.

steps
  1. Have a Secret Scanning project with a GitHub data source whose repository has been scanned 50 times (one full scan that found 5 leaks, then 49 pushes)
  2. Open the project's Data Sources list
  3. Open the data source and its Repositories list
  4. Count the rows the database returns for each list's read
the Data Sources dashboard reads about one row per data source shown
e2e/specs/secret-scanning/s13-dashboards-read-every-scan.spec.ts:59
lines
infisical/backend/src/ee/services/secret-scanning-v2/secret-scanning-v2-dal.ts:227
infisical/backend/src/ee/services/secret-scanning-v2/secret-scanning-v2-dal.ts:288
infisical/backend/src/ee/services/secret-scanning-v2/secret-scanning-v2-dal.ts:309
infisical/backend/src/ee/services/secret-scanning-v2/secret-scanning-v2-dal.ts:372
infisical/backend/src/ee/routes/v2/secret-scanning-v2-routers/secret-scanning-v2-router.ts:431
infisical/backend/src/ee/routes/v2/secret-scanning-v2-routers/secret-scanning-v2-endpoints.ts:528
infisical/frontend/src/pages/secret-scanning/SecretScanningDataSourcesPage/components/SecretScanningDataSourcesSection.tsx:28
infisical/frontend/src/pages/secret-scanning/SecretScanningDataSourceByIdPage/components/SecretScanningResourceTable.tsx:84
before the PRs testedno
A failed load of PAM sessions shows "No sessions found" instead of an error
s2-failed-session-list-shown-as-no-sessions-1
uiloss: decision
medium

When

Got

Want

steps
  1. Sign in as a member of an organisation with PAM on the enterprise plan
  2. Make the load of the session list fail (the server answers an error)
  3. Open PAM > Sessions
  4. Wait for the page to settle
lines
infisical/frontend/src/pages/pam/PamSessionsPage/PamSessionsPage.tsx:162
infisical/frontend/src/pages/pam/PamSessionsPage/PamSessionsPage.tsx:168
infisical/frontend/src/pages/pam/PamSessionsPage/PamSessionsPage.tsx:252
infisical/frontend/src/hooks/api/pam/queries.tsx:403
before the PRs testedno
A PAM folder loads every account it holds in one go, with no paging
s7-folder-account-list-loads-every-account-1
pamperformanceloss: time
medium

When You open a PAM folder that holds many privileged accounts.

Got Every account in the folder is loaded at once, so the folder gets slower the more accounts it holds.

Want The folder loads a page of accounts at a time.

steps
  1. Create a PAM folder and an AWS IAM account template.
  2. Create 150 AWS IAM accounts in that folder.
  3. Open the folder (its page loads the folder's account list once).
s7 a PAM folder's account list returns at most a page of accounts
e2e/specs/pam/s7-folder-account-list-loads-every-account.spec.ts:8
lines
infisical/backend/src/ee/routes/v1/pam-routers/pam-account-router.ts:397
infisical/backend/src/ee/routes/v1/pam-routers/pam-account-router.ts:411
infisical/backend/src/ee/services/pam-account/pam-account-service.ts:340
infisical/backend/src/ee/services/pam-account/pam-account-service.ts:362
infisical/backend/src/ee/services/pam-account/pam-account-dal.ts:221
infisical/frontend/src/pages/pam/PamAccountsPage/PamFolderPage/PamFolderPage.tsx:98
infisical/frontend/src/pages/pam/PamDiscoveryPage/components/DiscoveryConfigFields.tsx:271
before the PRs testedno
A failed folder load makes PAM look empty or the folder look deleted
s7-failed-folder-list-shown-as-no-accounts-1
pambehaviourloss: time
medium

When PAM folders cannot be loaded for a moment.

Got PAM Accounts says there are no accounts yet, and a folder's page says the folder no longer exists.

Want Both pages say the folders could not be loaded.

steps
  1. Sign in as a PAM admin of an organisation that has at least one PAM folder
  2. Make loading the PAM folder list fail (the server answers an error twice; the page retries once)
  3. Open PAM > Accounts
  4. Open the folder's own page by its link
lines
infisical/frontend/src/hooks/api/pam/queries.tsx:270
infisical/frontend/src/pages/pam/PamAccountsPage/PamAccountsPage.tsx:92
infisical/frontend/src/pages/pam/PamAccountsPage/PamAccountsPage.tsx:165
infisical/frontend/src/pages/pam/PamAccountsPage/PamFolderPage/PamFolderPage.tsx:94
infisical/frontend/src/pages/pam/PamAccountsPage/PamFolderPage/PamFolderPage.tsx:193
before the PRs testedno
Cancelled PAM access requests look like they still await review
s7-cancelled-request-shown-as-pending-review-1
pamuiloss: decision
medium

When A pending access request's account is deleted, which cancels the request.

Got The folder's approvals history still shows the request as Pending Review.

Want The request is shown as cancelled.

steps
  1. Create a PAM folder and an AWS IAM account in it from a template with the Requires Approval policy on
  2. Add yourself as the folder's approver on its Approvals tab
  3. Request access to the account (the request stays pending)
  4. Delete the account
  5. Open the folder's Approvals tab and read the Access Requests list
s7-cancelled-request-shown-as-pending-review
e2e/specs/pam/s7-cancelled-request-shown-as-pending-review.spec.ts:5
lines
infisical/frontend/src/pages/pam/components/approvalRequestStatus.ts:6
infisical/frontend/src/pages/pam/components/approvalRequestStatus.ts:28
infisical/frontend/src/pages/pam/PamAccountsPage/components/FolderApprovalsTab.tsx:649
infisical/backend/src/services/approval-policy/approval-policy-service.ts:2123
before the PRs testedno
Resolving many findings at once does the database work once per finding
s12-findings-bulk-update-statements-per-finding-1
secret-scanningperformanceloss: time
medium

When You select many findings and change their status together.

Got The product redoes the same work for each finding, so a larger selection takes proportionally longer.

Want The whole selection is updated in one go.

steps
  1. Sign in as a member of a Secret Scanning project whose repository has many unresolved findings
  2. Open Findings and select 5 findings, then Update Status to Resolved
  3. Select 50 other findings and Update Status to Resolved
  4. Compare the database statements each batch made
updating the status of 50 selected findings makes no more statements per finding than updating 5
e2e/specs/secret-scanning/s12-findings-bulk-update-statements-per-finding.spec.ts:42
lines
infisical/backend/src/ee/routes/v2/secret-scanning-v2-routers/secret-scanning-v2-router.ts:311
infisical/backend/src/ee/services/secret-scanning-v2/secret-scanning-v2-service.ts:805
infisical/backend/src/ee/services/secret-scanning-v2/secret-scanning-v2-service.ts:829
before the PRs testedno
The PAM discovery sources list gets slower with every scan a source has ever run
s8-discovery-sources-list-reads-every-run-1
pamperformanceloss: time
medium

When Your PAM discovery sources have been scanning for a while.

Got Opening the Discovery screen reads every scan each source ever ran, so it gets slower over time.

Want The screen reads only each source's latest scan.

steps
  1. Create a PAM discovery source and let it run many scans (a daily schedule, or Scan repeatedly)
  2. Open PAM > Discovery, which loads the discovery sources list
  3. Count the scan-run rows the database returns for that one load
s8 the discovery sources list reads one run per source
e2e/specs/pam/s8-discovery-sources-list-reads-every-run.spec.ts:17
lines
infisical/backend/src/ee/services/pam-discovery/pam-discovery-source-service.ts:283
infisical/backend/src/ee/services/pam-discovery/pam-discovery-source-service.ts:291
infisical/backend/src/ee/services/pam-discovery/pam-discovery-source-service.ts:310
infisical/backend/src/ee/routes/v1/pam-routers/pam-discovery-router.ts:311
before the PRs testedno
Pressing Esc on the Add AWS KMS form throws away what was typed, with no warning
s17-aws-kms-draft-lost-on-escape-1
kmsuiloss: time
low

When You start adding an AWS KMS, type an alias and description, and press Esc.

Got The form closes straight away with no warning. When you reopen it, everything you typed is gone.

Want You are asked before your input is thrown away, or it is still there when you reopen the form.

steps
  1. Sign in as an organisation admin on the enterprise license.
  2. Open Organization Settings, Encryption tab.
  3. Click Add KMS, choose AWS KMS.
  4. Type an alias and a description.
  5. Press Esc.
  6. Click Add KMS again and choose AWS KMS.
s17 — Esc on a half-filled Add AWS KMS form keeps the alias or warns first
e2e/specs/kms/s17-aws-kms-draft-lost-on-escape.spec.ts:17
lines
infisical/frontend/src/pages/organization/SettingsPage/components/OrgEncryptionTab/AddExternalKmsForm.tsx:53
infisical/frontend/src/pages/organization/SettingsPage/components/OrgEncryptionTab/AwsKmsForm.tsx:94
before the PRs testedno
Going Back in "Add External KMS" throws away the typed AWS KMS alias
s9-aws-kms-alias-lost-on-back-1
behaviourloss: time
low

When

Got

Want

steps
  1. Sign in as an organization admin on the enterprise license and open Organization Settings > Encryption.
  2. Click Add KMS, then pick AWS KMS.
  3. Type an Alias.
  4. Click Back.
  5. Pick AWS KMS again.
lines
infisical/frontend/src/pages/organization/SettingsPage/components/OrgEncryptionTab/AddExternalKmsForm.tsx:110
infisical/frontend/src/pages/organization/SettingsPage/components/OrgEncryptionTab/AwsKmsForm.tsx:95
before the PRs testedno
Approval request search misses requests that are not on the current page
s8-approval-search-misses-later-pages-1
pambehaviourloss: time
medium

When You have more than 20 access requests to approve and search for one.

Got If the request is not on the page you are viewing, the list says no requests match your search.

Want The search finds the request whichever page it was on.

steps
  1. Make a PAM folder whose approval policy names you as approver, with accounts from a template that requires approval
  2. Request access to 21 of those accounts so more than 20 requests await your approval
  3. Open PAM > Approval Requests and note the pager says 1 - 20 of 21 or more
  4. Type the account name of a request that is not on page 1 into the search
s8-approval-search-misses-later-pages
e2e/specs/pam/s8-approval-search-misses-later-pages.spec.ts:7
lines
infisical/frontend/src/pages/pam/PamApprovalRequestsPage/PamApprovalRequestsPage.tsx:213
infisical/frontend/src/pages/pam/PamApprovalRequestsPage/PamApprovalRequestsPage.tsx:103
infisical/frontend/src/pages/pam/PamApprovalRequestsPage/PamApprovalRequestsPage.tsx:265
before the PRs testedno
Pressing Escape in the Create CA panels closes them with no warning and discards what was typed
s4-create-ca-panel-esc-discards-draft-1
behaviourloss: time
low

When

Got

Want

steps
  1. Open a cert-manager project's Certificate Authorities page
  2. Click Create CA and type a name in the name field (placeholder my-internal-ca)
  3. Press Escape
  4. Click Create CA again: the name field is empty and no confirmation was shown
  5. On the External tab, click Create External CA, pick Let's Encrypt, type a name (placeholder my-external-ca), press Escape and reopen: the name is gone too
lines
infisical/frontend/src/pages/cert-manager/CertificateAuthoritiesPage/components/CreateCaWizard/CreateCaWizard.tsx:101
infisical/frontend/src/pages/cert-manager/CertificateAuthoritiesPage/components/ExternalCaModal.tsx:956
#8637
reviewed by greptile, cubic, coderabbit
no
Going back with "Select Another CA" and choosing ACME again empties the external CA name already typed
s4-acme-name-wiped-on-select-another-ca-1
behaviourloss: time
low

When

Got

Want

steps
  1. Open a cert-manager project's Certificate Authorities page on the External tab
  2. Click Create External CA and pick Let's Encrypt (ACME)
  3. Type a name in the name field (placeholder my-external-ca)
  4. Click "Select Another CA"
  5. Pick Let's Encrypt (ACME) again: the name field is empty
lines
infisical/frontend/src/pages/cert-manager/CertificateAuthoritiesPage/components/ExternalCaModal.tsx:370
infisical/frontend/src/pages/cert-manager/CertificateAuthoritiesPage/components/ExternalCaModal.tsx:978
before the PRs testedno
Searching for a CA type that matches nothing leaves the Connect External CA panel blank, with no "no results" message
s5-external-ca-type-search-no-match-blank-1
uiloss: nothing
low

When

Got

Want

steps
  1. Open a Certificate Management project's Certificate Authorities page on the External tab
  2. Click Create External CA
  3. Type "zzqx-no-such-ca" into "Search certificate authorities..."
  4. Every CA type card disappears and the area under the search box is empty — no message says nothing matched
lines
infisical/frontend/src/pages/cert-manager/CertificateAuthoritiesPage/components/ExternalCaModal.tsx:1010
infisical/frontend/src/pages/cert-manager/CertificateAuthoritiesPage/components/ExternalCaModal.tsx:940
before the PRs testedno
Cancelling a filled Add Key panel on a KMS project throws away the key name and description, with no warning
s8-add-key-cancel-discards-draft-1
behaviourloss: time
low

When

Got

Want

steps
  1. Sign in as an organisation admin and open a KMS project's Overview
  2. Click Add Key, then Create Key
  3. Type a key name and a description
  4. Click Cancel
  5. Click Add Key, then Create Key again
  6. The Name field is empty and the description is gone; no warning was shown on Cancel
lines
infisical/frontend/src/pages/kms/OverviewPage/components/CmekModal.tsx:314
infisical/frontend/src/pages/kms/OverviewPage/components/CmekModal.tsx:337
before the PRs testedno
A device re-enrolling over EST with an unreadable request gets a server error
s5-est-reenroll-malformed-csr-server-error-1
behaviourloss: time
low

When

Got

Want

steps
  1. Create a Certificate Management project with a root CA and an intermediate CA signed by it
  2. Create a certificate policy and an EST certificate profile on the intermediate with a known passphrase
  3. Enroll a device over EST (simpleenroll) with the passphrase and a valid CSR, keep the issued certificate
  4. Re-enroll (simplereenroll) with the passphrase, the issued certificate as client certificate, and a body that is a PEM block holding text that is not a PKCS#10 request
  5. Compare with enrolling (simpleenroll) with the same unreadable body
lines
infisical/backend/src/services/certificate-est-v3/certificate-est-v3-service.ts:387
infisical/backend/src/services/certificate-common/certificate-csr-utils.ts:44
before the PRs testedno
The Subscribers page shows an empty table when the subscriber list fails to load
s15-subscribers-load-failure-shown-as-empty-1
pki-enrollmentbehaviourloss: decision
medium

When You open the Subscribers page and the subscriber list can't be loaded.

Got You see an empty table with column headings and no message, as if the project had no subscribers.

Want The page tells you the subscribers could not be loaded.

steps
  1. Sign in as a project admin of a Certificate Management project that has a PKI subscriber.
  2. Make the subscriber list request fail with a server error (GET /api/v1/projects/:projectId/pki-subscribers answers 500).
  3. Open the project's Subscribers page.
  4. Wait for the read to give up.
s15-subscribers-load-failure-shown-as-empty
e2e/specs/pki-enrollment/s15-subscribers-load-failure-shown-as-empty.spec.ts:12
lines
infisical/frontend/src/pages/cert-manager/PkiSubscribersPage/components/PkiSubscribersTable.tsx:55
infisical/frontend/src/pages/cert-manager/PkiSubscribersPage/components/PkiSubscribersTable.tsx:69
before the PRs testedno
Submitting a new certificate profile twice at once fails with a server error
s13-duplicate-profile-name-race-crashes-1
certificatesbehaviourloss: nothing
low

When You click Create Profile twice quickly, or two people create the same profile name at once.

Got The profile is created once, but the other attempt fails with 'Something went wrong'.

Want The second attempt says a profile with this name already exists.

steps
  1. Sign in and open a Certificate Manager project that has a certificate policy.
  2. Create a self-signed certificate profile named X, then create X again: the second is refused because the name already exists.
  3. Send two creates of a self-signed profile with a new name Y so they reach the database at the same moment (a double click on Create Profile in the wizard does this).
  4. Read both answers and the project's profile list.
two creates of one certificate profile name at once: one created, the other refused as already existing
e2e/specs/certificates/s13-duplicate-profile-name-race-crashes.spec.ts:16
lines
infisical/backend/src/services/certificate-profile/certificate-profile-service.ts:528
infisical/backend/src/services/certificate-profile/certificate-profile-dal.ts:60
before the PRs testedno
Two simultaneous saves give a PAM folder two approval steps instead of one
s9-overlapping-approver-saves-leave-two-steps-1
pambehaviourloss: work
low

When Two admins save the same PAM folder's approvers at the same moment.

Got Both saves succeed and the folder now has two approval steps, though PAM folders allow only one.

Want The folder keeps one approval step, the last save's.

steps
  1. Create a PAM folder in an organisation with PAM on the enterprise plan
  2. Save the folder's approvers (one step, the signed-in user as approver) twice in turn: the folder shows one approval step
  3. Hold the policy's step row from another database session (select ... for update on approval_policy_steps by policyId), then send two one-step approver saves for the same folder
  4. Release the hold once both saves are waiting on it, and read the folder's approval configuration
two overlapping approver saves leave the PAM folder with one approval step
e2e/specs/pam/s9-overlapping-approver-saves-leave-two-steps.spec.ts:10
lines
infisical/backend/src/services/approval-policy/approval-policy-service.ts:876
infisical/backend/src/services/approval-policy/approval-policy-service.ts:879
infisical/backend/src/services/approval-policy/approval-policy-service.ts:883
infisical/backend/src/ee/services/pam-access-request/pam-access-request-service.ts:199
infisical/backend/src/db/migrations/20251203002657_global-approvals.ts:39
before the PRs testedno
PKI subscribers can't be edited or disabled from the Subscribers page
s15-subscriber-edit-refused-missing-project-1
pki-enrollmentbehaviourloss: work
high

When You edit or disable a PKI subscriber from the Subscribers page.

Got Saving is refused with a validation error about a missing project, and the subscriber stays unchanged.

Want The change is saved and shown in the list.

steps
  1. Sign in as a project admin of a Certificate Management project that has an active internal CA.
  2. Create a PKI subscriber with the common name heal.example.com.
  3. Open the project's Subscribers page.
  4. Open the subscriber's row menu and choose Edit Subscriber.
  5. Change the common name to edited.heal.example.com and click Update.
lines
infisical/frontend/src/hooks/api/pkiSubscriber/mutations.tsx:41
infisical/backend/src/server/routes/v1/pki-subscriber-router.ts:259
infisical/frontend/src/config/request.ts:41
infisical/frontend/src/pages/cert-manager/PkiSubscribersPage/components/PkiSubscriberModal.tsx:341
infisical/frontend/src/pages/cert-manager/PkiSubscribersPage/components/PkiSubscriberSection.tsx:54
before the PRs testedno
Two SSH CA setups of one PAM account at once hand out two different CA keys
s9-ssh-ca-setups-at-once-create-two-cas-1
pambehaviourloss: time
medium

When Two SSH CA setups for one new PAM SSH account run at the same moment.

Got Each gets a different CA key. The account keeps only one, so one server can end up trusting a key PAM no longer signs with.

Want Both get the same CA key, the one the account keeps.

steps
  1. Create a PAM SSH account (certificate auth) in a folder; it has no certificate authority yet
  2. Run the account's SSH CA install command (or ask for its SSH CA public key) from two places at the same moment
  3. Compare the CA public keys the two answers carry, and the one the account keeps afterwards
s9 two SSH CA setups of one PAM account at once hand out one CA key
e2e/specs/pam/s9-ssh-ca-setups-at-once-create-two-cas.spec.ts:26
lines
infisical/backend/src/ee/services/pam-account/pam-account-service.ts:1365
infisical/backend/src/ee/services/pam-account/pam-account-service.ts:1377
infisical/backend/src/ee/services/pam-account/pam-account-service.ts:1390
infisical/backend/src/ee/routes/v1/pam-routers/pam-account-router.ts:1042
infisical/frontend/src/pages/pam/PamAccountsPage/components/SshCaSetupCallout.tsx:12
before the PRs testedno
A failed load of PAM requests awaiting my approval shows "No requests found" instead of an error
s3-failed-pending-approvals-shown-as-no-requests-1
uiloss: decision
medium

When

Got

Want

steps
  1. Sign in as a PAM approver in an organisation on the enterprise plan
  2. Make the load of requests awaiting approval fail (the server answers an error), including its one automatic retry
  3. Open PAM > Approval Requests
  4. Wait for the page to settle
lines
infisical/frontend/src/hooks/api/pam/queries.tsx:664
infisical/frontend/src/pages/pam/PamApprovalRequestsPage/PamApprovalRequestsPage.tsx:172
infisical/frontend/src/pages/pam/PamApprovalRequestsPage/PamApprovalRequestsPage.tsx:208
infisical/frontend/src/pages/pam/PamApprovalRequestsPage/PamApprovalRequestsPage.tsx:111
before the PRs testedno
Two access requests for one PAM account sent together both stay pending
s9-concurrent-access-requests-both-pending-1
pambehaviourloss: time
low

When You send two access requests for the same PAM account at the same moment, for example from two tabs.

Got Both are accepted, and the approvers have the same request waiting twice.

Want One is kept and the other is refused because you already have a pending request.

steps
  1. Create a PAM folder and set its approval policy with yourself as approver.
  2. Create an account template that requires approval and an AWS IAM account in the folder using it.
  3. Request access to the account, then request it again: the second is refused with 'You already have a pending request for this account'.
  4. On a second account of the folder, send two access requests for it at the same moment.
  5. Count the pending access requests for that second account.
a PAM account keeps one pending access request when two requests for it arrive together
e2e/specs/pam/s9-concurrent-access-requests-both-pending.spec.ts:10
lines
infisical/backend/src/ee/services/pam-access-request/pam-access-approval-resource.ts:470
infisical/backend/src/ee/services/pam-access-request/pam-access-approval-resource.ts:489
infisical/backend/src/services/approval-policy/approval-policy-service.ts:1104
infisical/backend/src/services/approval-policy/approval-policy-service.ts:1129
infisical/backend/src/ee/services/pam-access-request/pam-access-request-service.ts:414
before the PRs testedno
Two saves of a PAM folder's Slack notifications at once leave it with two copies
s10-overlapping-slack-notification-saves-keep-both-1
pambehaviourloss: time
low

When Two people save a PAM folder's Slack notification settings at the same moment.

Got Both saves succeed and the folder now has the same Slack notification twice.

Want The folder keeps one copy of the notification settings.

steps
  1. In an organisation with a Slack workflow integration and the PAM Slack notifications plan feature, create a PAM folder.
  2. Save the folder's approval configuration with one Slack notification (one channel, event access-requested); save it again — the folder lists one notification.
  3. Send the same save twice so that both transactions overlap on the folder's notification rows (forced in the test by holding a row lock on the folder's pam_folder_notification_configs row until both saves are waiting).
  4. Read the folder's approval configuration.
two overlapping Slack notification saves leave the PAM folder with one notification config
e2e/specs/pam/s10-overlapping-slack-notification-saves-keep-both.spec.ts:10
lines
infisical/backend/src/ee/services/pam-access-request/pam-access-request-service.ts:249
infisical/backend/src/ee/services/pam-access-request/pam-access-request-service.ts:250
infisical/backend/src/ee/services/pam-access-request/pam-access-request-service.ts:252
infisical/backend/src/db/migrations/20260708143943_pam-folder-notification-configs.ts:13
#8462
reviewed by greptile, cubic, coderabbit
no
A refused PAM launch still shows up as an active session
s10-refused-launch-leaves-active-session-1
pambehaviourloss: decision
medium

When You launch a session on a PAM account whose gateway is not set up yet.

Got The launch fails, but PAM > Sessions shows an active session for you on that account for the next hour.

Want A failed launch leaves no active session behind.

steps
  1. As an org admin, create a gateway (PAM gateway with token auth) and do not connect it
  2. Create a PostgreSQL account template and a folder in PAM
  3. Create a PostgreSQL account in the folder bound to that gateway (creation succeeds: the connection test is skipped for an unreachable gateway)
  4. Launch a session on the account by its path folder/account (POST /api/v1/pam/accounts/access, as the CLI does)
  5. The launch is refused with 'Gateway Config for org ... not found.'
  6. Open PAM > Sessions and search for the account
a PAM launch refused at the gateway step leaves no active session
e2e/specs/pam/s10-refused-launch-leaves-active-session.spec.ts:5
lines
infisical/backend/src/ee/services/pam-session/pam-session-service.ts:710
infisical/backend/src/ee/services/pam-session/pam-session-service.ts:740
infisical/backend/src/ee/services/pam-session/pam-session-service.ts:741
infisical/backend/src/ee/services/pam-session/pam-session-service.ts:743
infisical/backend/src/ee/services/gateway-v2/gateway-v2-service.ts:555
infisical/backend/src/ee/services/gateway-v2/gateway-v2-service.ts:565
infisical/backend/src/ee/services/gateway-v2/gateway-v2-service.ts:570
before the PRs testedno
Access Bundles says there are none when the list fails to load
s9-bundle-list-failure-shown-as-no-bundles-1
agent-vaultuiloss: decision
medium

When You open Access Bundles while the list cannot be loaded.

Got The page says you have no access bundles yet, with no sign anything went wrong.

Want The page tells you the bundles could not be loaded.

steps
  1. Sign in as an org admin whose organisation has at least one Agent Vault access bundle
  2. Make the access-bundle list request fail with a server error (both the first try and the automatic retry)
  3. Open Agent Vault, then Access Bundles
s9-bundle-list-failure-shown-as-no-bundles
e2e/specs/agent-vault/s9-bundle-list-failure-shown-as-no-bundles.spec.ts:10
lines
infisical/frontend/src/pages/agent-vault/AgentVaultAccessBundlesPage/AgentVaultAccessBundlesPage.tsx:100
infisical/frontend/src/pages/agent-vault/AgentVaultAccessBundlesPage/AgentVaultAccessBundlesPage.tsx:193
infisical/frontend/src/hooks/api/agentVault/queries.tsx:201
before the PRs testedno
The PAM approvals badge loads every past access request just to count pending ones
s8-pending-approval-count-reads-decided-requests-1
pamperformanceloss: time
medium

When Your organisation's PAM has built up many approved and rejected access requests.

Got Every time the approvals badge or list refreshes, every past request is loaded just to find the few still waiting, so it slows as history grows.

Want The badge costs the same with thousands of past requests as with five.

steps
  1. Create a PAM folder with an approval policy naming yourself, and an AWS IAM account in it whose template requires approval.
  2. File 50 access requests to the account and reject each one.
  3. Ask for your pending-approval count and note the access request rows the database returned for it.
  4. Compare those rows with the number of pending access requests in the organisation's PAM project.
lines
infisical/backend/src/services/approval-policy/approval-policy-service.ts:2224
infisical/backend/src/services/approval-policy/approval-request-dal.ts:91
infisical/backend/src/services/approval-policy/approval-request-dal.ts:117
infisical/backend/src/ee/services/pam-access-request/pam-access-request-service.ts:499
infisical/frontend/src/hooks/api/pam/queries.tsx:687
before the PRs testedno
A signer's Approvals tab spins forever, with no error, when its approval policy fails to load
s4-failed-approval-policy-load-never-reported-1
behaviourloss: time
medium

When

Got

Want

steps
  1. In a Certificate Manager project, import a code-signing certificate and create a code signer on it
  2. Give the signer an approval policy of one step (the admin as approver, max 5 signings)
  3. Make the signer's approval-policy request fail with a server error
  4. Open the signer's page on the Approvals tab and wait past the query's one retry
lines
infisical/frontend/src/pages/cert-manager/SignerDetailPage/components/SignerApprovalsSection.tsx:18
infisical/frontend/src/pages/cert-manager/SignerDetailPage/components/SignerApprovalsSection.tsx:25
infisical/frontend/src/pages/cert-manager/SignerDetailPage/components/SignerApprovalPolicyTab.tsx:123
infisical/frontend/src/hooks/api/signers/queries.tsx:183
before the PRs testedno
Agent Vault Proxies page shows "No proxies yet" when the proxy list fails to load
s1-proxies-load-failure-shown-as-no-proxies-1
behaviourloss: decision
medium

When

Got

Want

steps
  1. Sign in as an org admin with Agent Vault available.
  2. Make the proxy list request fail with a server error (500), including the automatic retry.
  3. Open Agent Vault > Proxies.
  4. The page shows the empty state "No proxies yet — Create a proxy and enroll it where agent traffic leaves your network." and no error.
lines
infisical/frontend/src/pages/agent-vault/AgentVaultProxiesPage/AgentVaultProxiesPage.tsx:151
infisical/frontend/src/pages/agent-vault/AgentVaultProxiesPage/AgentVaultProxiesPage.tsx:159
infisical/frontend/src/pages/agent-vault/AgentVaultProxiesPage/AgentVaultProxiesPage.tsx:254
infisical/frontend/src/hooks/api/agentVault/queries.tsx:252
before the PRs testedno
Approval requests load slower with every folder the approver oversees
s7-pending-approvals-checked-per-folder-1
pamperformanceloss: time
medium

When You approve PAM access for many folders and open Approval Requests.

Got The page does more work for every folder with a pending request, so it loads slower as folders are added.

Want Approval Requests loads in the same time however many folders you oversee.

steps
  1. As a PAM approver, set yourself as approver on 2 folders whose accounts' template requires approval, and request access to one account in each
  2. Add 8 more such folders, each with one pending request
  3. Open PAM > Approval Requests and count the database statements one load makes
lines
infisical/backend/src/services/approval-policy/approval-policy-service.ts:2236
infisical/backend/src/ee/services/pam-access-request/pam-access-approval-resource.ts:856
infisical/backend/src/ee/services/pam-access-request/pam-access-approval-resource.ts:215
infisical/backend/src/ee/services/pam-access-request/pam-access-approval-resource.ts:177
#8462
reviewed by greptile, cubic, coderabbit
no
Sessions says there are none when the session list fails to load
s10-session-list-failure-shown-as-no-sessions-1
agent-vaultuiloss: decision
medium

When You open Sessions while the session list cannot be loaded.

Got The page says you have no sessions yet, with no sign anything went wrong.

Want The page tells you the sessions could not be loaded.

steps
  1. Sign in as an org admin whose organisation has an access bundle and an active Agent Vault session on it
  2. Make the session list request fail with a server error (both the first try and the automatic retry)
  3. Open Agent Vault, then Sessions
s10-session-list-failure-shown-as-no-sessions
e2e/specs/agent-vault/s10-session-list-failure-shown-as-no-sessions.spec.ts:10
lines
infisical/frontend/src/pages/agent-vault/AgentVaultSessionsPage/AgentVaultSessionsPage.tsx:135
infisical/frontend/src/pages/agent-vault/AgentVaultSessionsPage/AgentVaultSessionsPage.tsx:350
infisical/frontend/src/hooks/api/agentVault/queries.tsx:232
before the PRs testedno
An access bundle that fails to load is reported as deleted or no longer granted
s2-bundle-load-failure-shown-as-deleted-1
uiloss: decision
medium

When

Got

Want

steps
  1. Sign in as an org admin and create an Agent Vault access bundle
  2. Make the bundle's detail request fail with a server error (both the first try and the automatic retry)
  3. Open that access bundle's detail page
lines
infisical/frontend/src/pages/agent-vault/AgentVaultAccessBundleDetailPage/AgentVaultAccessBundleDetailPage.tsx:52
infisical/frontend/src/pages/agent-vault/AgentVaultAccessBundleDetailPage/AgentVaultAccessBundleDetailPage.tsx:61
infisical/frontend/src/hooks/api/agentVault/queries.tsx:217
before the PRs testedno
An over-long member search says no members match instead of reporting the error
s11-long-member-search-shown-as-no-match-1
agent-vaultbehaviourloss: nothing
low

When You search an access bundle's members with a very long search (over 255 characters).

Got The sheet says no members match and to try a different term, though the search was never run.

Want The sheet tells you the search is too long, or stops you typing past the limit.

steps
  1. Sign in as an org admin with Agent Vault available and create an access bundle
  2. Open the bundle's detail page, then Options › Manage Access
  3. Type or paste a search of 256 characters into Search members...
s11-long-member-search-shown-as-no-match
e2e/specs/agent-vault/s11-long-member-search-shown-as-no-match.spec.ts:9
lines
infisical/backend/src/ee/routes/v1/agent-vault-routers/agent-vault-schemas.ts:416
infisical/backend/src/ee/routes/v1/agent-vault-routers/agent-vault-access-bundle-router.ts:760
infisical/frontend/src/components/agent-vault/ManageAccessSheet.tsx:80
infisical/frontend/src/components/agent-vault/ManageAccessSheet.tsx:134
infisical/frontend/src/components/agent-vault/ManageAccessSheet.tsx:149
before the PRs testedno
The Sessions list re-reads a member's whole session history to count the active ones
s11-active-session-count-walks-session-history-1
agent-vaultperformanceloss: time
medium

When You filter your Agent Vault sessions to Active after many sessions have come and gone.

Got Each load and each 30-second refresh reads every session you ever had, so the page gets slower as your history grows.

Want Showing your few active sessions costs the same however long your history is.

steps
  1. Sign in as a user in their own organisation with Agent Vault available, and create an access bundle
  2. Create 3 agent sessions on that bundle with a 1-hour expiry
  3. Have the same user hold a history of 20,000 revoked agent sessions in that Agent Vault (each session the product mints and revokes adds one)
  4. List the user's own sessions filtered to Active (GET /api/v1/agent-vault/sessions?scope=mine&status=active)
  5. Run EXPLAIN (ANALYZE) on the count statement the list ran
s11-active-session-count-walks-session-history
e2e/specs/agent-vault/s11-active-session-count-walks-session-history.spec.ts:17
lines
infisical/backend/src/ee/services/agent-vault-session/agent-vault-session-dal.ts:49
infisical/backend/src/ee/services/agent-vault-session/agent-vault-session-dal.ts:139
infisical/backend/src/ee/services/agent-vault-session/agent-vault-session-dal.ts:195
infisical/backend/src/db/migrations/20260907100804_agent-vault.ts:78
infisical/backend/src/db/migrations/20260907100804_agent-vault.ts:96
infisical/frontend/src/pages/agent-vault/AgentVaultSessionsPage/AgentVaultSessionsPage.tsx:135
before the PRs testedno
An access bundle accepts any number of services and opening it loads them all
s11-bundle-services-unbounded-1
agent-vaultperformanceloss: time
low

When You keep adding services to one access bundle.

Got Every service is accepted, and opening the bundle loads all of them, with their headers and substitutions, every time.

Want The bundle shows its services a page at a time, or stops you at a set limit like it does for variables.

steps
  1. As an org admin, create an access bundle in Agent Vault
  2. Add 156 services to it, each with its own host
  3. Open the access bundle
s11-bundle-services-unbounded
e2e/specs/agent-vault/s11-bundle-services-unbounded.spec.ts:9
lines
infisical/backend/src/ee/services/agent-vault-access-bundle/agent-vault-service-dal.ts:15
infisical/backend/src/ee/services/agent-vault-access-bundle/agent-vault-access-bundle-service.ts:661
infisical/backend/src/ee/services/agent-vault-access-bundle/agent-vault-access-bundle-service.ts:787
infisical/backend/src/ee/services/agent-vault/agent-vault-variable-fns.ts:5
before the PRs testedno
A person removed from Agent Vault can keep an access bundle and get it back on re-add
s12-removed-member-keeps-bundle-grant-1
agent-vaultbehaviourloss: decision
medium

When An admin removes someone from Agent Vault while another admin is granting them an access bundle.

Got The removal says done, but the person stays listed on the bundle, and when re-added later they have the bundle back.

Want Removing someone from Agent Vault takes every bundle they hold, including one granted at the same moment.

steps
  1. Invite a person to the org with Agent Vault access, and create an access bundle
  2. Start granting the bundle to that person; hold the grant at the bundle row lock (after its membership check)
  3. While the grant waits, remove the person from Agent Vault — the removal answers 200
  4. Let the grant finish — it answers 200
  5. Open the bundle's member list; then add the person back to Agent Vault and open it again
s12-removed-member-keeps-bundle-grant: removed member holds no bundle
e2e/specs/agent-vault/s12-removed-member-keeps-bundle-grant.spec.ts:50
lines
infisical/backend/src/ee/services/agent-vault-access-bundle/agent-vault-access-bundle-service.ts:1576
infisical/backend/src/ee/services/agent-vault-access-bundle/agent-vault-access-bundle-service.ts:1584
infisical/backend/src/ee/services/agent-vault-member/agent-vault-membership-service.ts:619
infisical/backend/src/ee/services/agent-vault-member/agent-vault-membership-service.ts:635
before the PRs testedno
Sorting access bundles by services gets slower with every org's services
s12-bundle-list-by-services-reads-every-org-1
agent-vaultperformanceloss: time
medium

When You sort an organisation's Access Bundles by number of services.

Got Each page counts the services of every organisation on the instance, so it slows down as other organisations add services.

Want The page only looks at your organisation's services.

steps
  1. Sign in as an org admin and create a second organisation; in it create one access bundle and add 2 services
  2. In the first organisation create an access bundle and add 150 services
  3. In the second organisation open Agent Vault Access Bundles and sort by Services (the list asks for orderBy=serviceCount)
  4. Read the plan of the page statement the product ran (EXPLAIN ANALYZE of the captured statement, with this org's project id and limit 1)
lines
infisical/backend/src/ee/services/agent-vault-access-bundle/agent-vault-access-bundle-dal.ts:94
infisical/backend/src/ee/services/agent-vault-access-bundle/agent-vault-access-bundle-dal.ts:101
infisical/backend/src/ee/services/agent-vault-access-bundle/agent-vault-access-bundle-dal.ts:109
before the PRs testedno
An application that fails to load is shown as not existing
s8-failed-application-lookup-shown-as-not-found-1
behaviourloss: decision
medium

When

Got

Want

steps
  1. In the org's active Certificate Manager project, create an application
  2. Make the application lookup fail on the server (it answers a server error)
  3. Open the application's page under Certificate Manager > Applications
lines
infisical/frontend/src/pages/cert-manager/ApplicationDetailsByIDPage/ApplicationDetailsByIDPage.tsx:182
infisical/frontend/src/pages/cert-manager/ApplicationDetailsByIDPage/ApplicationDetailsByIDPage.tsx:170
before the PRs testedno
A relay list that fails to load is shown as "No relays configured" instead of an error
s2-failed-relay-list-shown-as-no-relays-1
uiloss: decision
medium

When

Got

Want

steps
  1. Sign in as an organisation admin.
  2. Make the relay list request fail with a server error.
  3. Open Organization Settings > Networking and select the Relays tab.
  4. The tab shows the empty state "No relays configured" and no error.
lines
infisical/frontend/src/pages/organization/NetworkingPage/components/RelayTab/RelayTab.tsx:105
infisical/frontend/src/pages/organization/NetworkingPage/components/RelayTab/RelayTab.tsx:179
infisical/frontend/src/hooks/api/relays/queries.tsx:31
before the PRs testedno
A gateway whose connected resources fail to load says "No connected resources" instead of an error
s4-failed-gateway-resources-shown-as-none-1
uiloss: decision
medium

When

Got

Want

steps
  1. Sign in as an organisation admin and register a gateway
  2. Make the gateway's connected-resources request fail with a server error
  3. Open Networking, then the gateway's detail page
  4. Look at the Connected Resources card
lines
infisical/frontend/src/pages/organization/NetworkingPage/GatewayDetailsByIDPage/components/GatewayConnectedResourcesSection/GatewayConnectedResourcesSection.tsx:74
infisical/frontend/src/pages/organization/NetworkingPage/GatewayDetailsByIDPage/components/GatewayConnectedResourcesSection/GatewayConnectedResourcesSection.tsx:95
before the PRs testedno
Create Application reopens on a draft the user cancelled
s6-create-application-reopens-cancelled-draft-1
behaviourloss: nothing
low

When

Got

Want

steps
  1. Open a Certificate Management project's Applications page
  2. Click Create Application, type a name and a description
  3. Click Cancel
  4. Click Create Application again
lines
infisical/frontend/src/pages/cert-manager/ApplicationsPage/components/PkiApplicationModal.tsx:82
infisical/frontend/src/pages/cert-manager/ApplicationsPage/components/PkiApplicationModal.tsx:213
before the PRs testedno
A gateway or relay detail page whose load fails says "Gateway not found" / "Relay not found" instead of an error
s5-failed-detail-load-shown-as-not-found-1
behaviourloss: decision
medium

When

Got

Want

steps
  1. Sign in as an organization admin and register a gateway (token auth) and a relay in the organization
  2. Make the server fail the gateway's detail read (it answers 500)
  3. Open Organization Settings > Networking and open the gateway's detail page
  4. The page says 'Gateway not found' and shows no error
  5. Do the same for the relay's detail read and page: it says 'Relay not found' and shows no error
lines
infisical/frontend/src/pages/organization/NetworkingPage/GatewayDetailsByIDPage/GatewayDetailsByIDPage.tsx:26
infisical/frontend/src/pages/organization/NetworkingPage/GatewayDetailsByIDPage/GatewayDetailsByIDPage.tsx:32
infisical/frontend/src/pages/organization/NetworkingPage/RelayDetailsByIDPage/RelayDetailsByIDPage.tsx:26
infisical/frontend/src/pages/organization/NetworkingPage/RelayDetailsByIDPage/RelayDetailsByIDPage.tsx:37
before the PRs testedno
A gateway list that fails to load shows "No gateways configured" instead of an error
s1-gateway-list-failure-shown-as-no-gateways-1
behaviourloss: decision
medium

When

Got

Want

steps
  1. Sign in as an organisation admin whose organisation has at least one gateway registered
  2. Make the gateway list request fail with a server error (GET /api/v2/gateways answers 500)
  3. Open Organization Settings > Networking, Gateways tab, All Gateways
  4. Wait for the list to finish loading (the request is retried once and fails again)
lines
infisical/frontend/src/pages/organization/NetworkingPage/components/GatewayTab/GatewayTab.tsx:113
infisical/frontend/src/pages/organization/NetworkingPage/components/GatewayTab/GatewayTab.tsx:272
infisical/frontend/src/hooks/api/gateways/queries.tsx:10
before the PRs testedno
A gateway pool list that fails to load shows "No gateway pools configured" instead of an error
s3-failed-pool-list-shown-as-no-pools-1
behaviourloss: decision
medium

When

Got

Want

steps
  1. As an organisation admin with gateway pools enabled, create a gateway pool
  2. Make the gateway pool list request fail with a server error
  3. Open Organization Settings > Networking and switch the Gateways card to Gateway Pools
  4. Observe the pool list
lines
infisical/frontend/src/pages/organization/NetworkingPage/components/GatewayTab/components/GatewayPoolsContent.tsx:62
infisical/frontend/src/pages/organization/NetworkingPage/components/GatewayTab/components/GatewayPoolsContent.tsx:100
infisical/frontend/src/hooks/api/gateway-pools/queries.tsx:30
before the PRs testedno
A discovery job whose details fail to load shows a blank page with no error
s6-failed-discovery-job-details-blank-1
uiloss: time
medium

When

Got

Want

steps
  1. Create a certificate discovery job in a Certificate Manager project
  2. Have the job's details read fail on the server (500)
  3. Open the job's details page under Certificate Manager > Discovery
  4. After the app's retry also fails, the main area is empty: no job name, no error, no toast, no alert
lines
infisical/frontend/src/pages/cert-manager/DiscoveryDetailsByIDPage/DiscoveryDetailsByIDPage.tsx:55
infisical/frontend/src/pages/cert-manager/DiscoveryDetailsByIDPage/DiscoveryDetailsByIDPage.tsx:82
before the PRs testedno
A data source vanishes from the Data Sources list once its leaks are resolved
s14-resolved-data-source-missing-from-list-1
secret-scanningbehaviourloss: decision
medium

When You resolve every leak a data source's scan found.

Got The data source disappears from the Data Sources list, as if it had been removed.

Want It stays in the list, showing no open leaks.

steps
  1. In a Secret Scanning project, have a GitHub data source whose repository was scanned once and found one leak
  2. Open the project's Data Sources page and search the data source's name: it is listed with 1 open leak
  3. Open Findings and mark that leak Resolved
  4. Go back to Data Sources, reload and search the data source's name again
a data source with all its findings resolved is still listed
e2e/specs/secret-scanning/s14-resolved-data-source-missing-from-list.spec.ts:6
lines
infisical/backend/src/ee/services/secret-scanning-v2/secret-scanning-v2-dal.ts:246
infisical/backend/src/ee/services/secret-scanning-v2/secret-scanning-v2-service.ts:134
infisical/backend/src/ee/services/secret-scanning-v2/secret-scanning-v2-dal.ts:309
before the PRs testedno
Certificate filter on an application's page announces its state on the wrong element
s19-application-certificate-filter-invalid-aria-1
pki-syncs-alertsuiloss: nothing
low

When You use a screen reader on an application's certificates list.

Got The filter's open or closed state is attached to an invisible wrapper, not the filter button you reach.

Want The filter button itself says whether its filter panel is open.

steps
  1. Sign in as a member of an organization with an active Certificate Manager project.
  2. Create an application, then open Certificate Manager > Applications > the application.
  3. Wait for the certificates list with its filter button.
  4. Run an accessibility check (axe) on the filter control.
s19-application-certificate-filter-invalid-aria
e2e/specs/pki-syncs-alerts/s19-application-certificate-filter-invalid-aria.spec.ts:7
lines
infisical/frontend/src/pages/cert-manager/CertificatesPage/components/CertificatesTable.tsx:753
infisical/frontend/src/components/v3/generic/Popover/Popover.tsx:14
before the PRs testedno
Pressing Escape in the Create Profile wizard closes it with no warning and discards the typed name and description
s8-create-profile-escape-discards-typed-input-1
behaviourloss: time
low

When

Got

Want

steps
  1. Open a Certificate Manager project's Certificate Profiles screen
  2. Open Create Profile and type a profile name (placeholder your-profile-name) and a description
  3. Press Escape
  4. Open Create Profile again: no confirmation was shown, and the name field is empty
lines
infisical/frontend/src/pages/cert-manager/PoliciesPage/components/CertificateProfilesTab/CreateProfileModal.tsx:935
infisical/frontend/src/pages/cert-manager/PoliciesPage/components/CertificateProfilesTab/CreateProfileModal.tsx:976
before the PRs testedno
A machine identity can end up with two relays though only one is allowed
s15-machine-identity-registers-two-relays-at-once-1
gateways-relaysbehaviourloss: nothing
low

When Two relay processes start with the same machine identity at the same moment.

Got Both are registered, so the identity now has two relays.

Want Only one relay is registered and the other start is refused.

steps
  1. As an organisation admin, create a machine identity with the admin role and universal auth, and log it in.
  2. As that identity, register relay R1 and then R2 one after the other: the second is refused because the identity already has a relay.
  3. With a fresh identity, send two relay registrations with different names at the same moment (forced in the test by holding the identity's row so both pass the existing-relay check before either inserts).
  4. Count the relays owned by that identity.
s15-machine-identity-registers-two-relays-at-once
e2e/specs/gateways-relays/s15-machine-identity-registers-two-relays-at-once.spec.ts:12
lines
infisical/backend/src/ee/services/relay/relay-service.ts:992
infisical/backend/src/ee/services/relay/relay-service.ts:1002
infisical/backend/src/ee/services/relay/relay-service.ts:1007
infisical/backend/src/db/migrations/20250825131627_add-gateway-v2-pki-and-ssh-configs.ts:98
before the PRs testedno
Showing what is connected to a gateway or gateway pool reads every organisation's app connections, dynamic secrets, Kubernetes auths and discovery configs
s11-gateway-resources-read-every-organisation-1
gateways-relaysperformanceloss: time
medium

When You open Networking or a gateway's or pool's connected resources on an instance with many app connections, dynamic secrets, Kubernetes auths or discovery configs.

Got Finding what is connected to the gateway looks through every one of those resources in every organisation, so the page gets slower as the whole instance grows.

Want Only the gateway's own connected resources are looked at, however big the instance gets.

steps
  1. Sign in as an organisation admin with at least one gateway
  2. Open Networking (the gateway list counts each gateway's connected resources) or open a gateway's detail page (GET /api/v2/gateways/:id/resources)
  3. Look at the statements the read runs: app_connections / dynamic_secrets / identity_kubernetes_auths / pki_discovery_configs filtered by their gatewayId, gatewayV2Id or gatewayPoolId column
  4. Look at the index catalog of those four tables
s11-gateway-resources-read-every-organisation
e2e/specs/gateways-relays/s11-gateway-resources-read-every-organisation.spec.ts:13
lines
infisical/backend/src/db/migrations/20250717195959_gatewayid-for-app-conn.ts:8
infisical/backend/src/db/migrations/20250901091637_add-gateway-v2-id-columns.ts:8
infisical/backend/src/db/migrations/20250901091637_add-gateway-v2-id-columns.ts:15
infisical/backend/src/db/migrations/20260205012343_add-pki-discovery-tables.ts:53
infisical/backend/src/db/migrations/20260501093623_add-gateway-pool-to-app-connections.ts:9
infisical/backend/src/db/migrations/20260501094157_add-gateway-pool-to-dynamic-secrets.ts:9
infisical/backend/src/db/migrations/20260501091201_add-gateway-pool-to-pki-discovery-configs.ts:9
infisical/backend/src/services/app-connection/app-connection-dal.ts:158
infisical/backend/src/ee/services/dynamic-secret/dynamic-secret-dal.ts:226
infisical/backend/src/ee/services/gateway-v2/gateway-v2-service.ts:334
infisical/backend/src/ee/services/gateway-v2/gateway-v2-service.ts:1294
infisical/backend/src/ee/services/gateway-pool/gateway-pool-service.ts:122
before the PRs testedno
Two enrollment tokens generated at once both stay valid
s11-concurrent-enrollment-tokens-both-enroll-1
gateways-relaysbehaviourloss: decision
low

When Two people generate a new enrollment token for the same relay or gateway at the same moment.

Got Both new tokens work, so a token you think was replaced can still enroll a relay or gateway.

Want Only the newest token works; every earlier one is refused.

steps
  1. Sign in as an organisation admin and create a relay (or gateway) with token authentication.
  2. Generate an enrollment token, then generate another: the first token is refused at relay login.
  3. Generate two enrollment tokens for the same relay at the same moment (the test forces the order by holding the relay's auth-method row so both requests delete before either inserts).
  4. Log in with each of the two new tokens.
s11-concurrent-enrollment-tokens-both-enroll
e2e/specs/gateways-relays/s11-concurrent-enrollment-tokens-both-enroll.spec.ts:10
lines
infisical/backend/src/ee/services/resource-auth-method/resource-auth-method-service.ts:1087
infisical/backend/src/ee/services/resource-auth-method/resource-auth-method-service.ts:1088
infisical/backend/src/ee/services/resource-auth-method/resource-auth-method-service.ts:1089
infisical/backend/src/db/migrations/20260430143000_resource-auth-methods.ts:115
before the PRs testedno
A gateway update reported as refused still switches its login method
s16-refused-gateway-rename-still-switches-auth-1
gateways-relaysbehaviourloss: decision
low

When You rename a gateway and change its login method in one save, while a colleague gives another gateway that name.

Got You are told the name already exists, but the login method changed anyway and the gateway's enrollment token is gone.

Want A refused save changes nothing, so the gateway keeps its login method and token.

steps
  1. In one organisation, create gateways A and C with token authentication and generate an enrollment token for A.
  2. Send an update for A that renames it to a new name N and switches it to AWS authentication (allowed account 123456789012).
  3. While A's rename is waiting to be saved (the test holds A's gateways_v2 row), rename C to N; it succeeds.
  4. Let A's update finish; it answers 400 "A gateway named N already exists".
  5. Read A's auth method and its enrollment tokens.
a gateway update refused for a name taken meanwhile leaves its token auth in place
e2e/specs/gateways-relays/s16-refused-gateway-rename-still-switches-auth.spec.ts:13
lines
infisical/backend/src/ee/routes/v3/gateway-router.ts:337
infisical/backend/src/ee/routes/v3/gateway-router.ts:348
infisical/backend/src/ee/routes/v3/gateway-router.ts:392
infisical/backend/src/ee/services/gateway-v2/gateway-v2-service.ts:1362
infisical/backend/src/ee/services/gateway-v2/gateway-v2-service.ts:1381
infisical/backend/src/ee/services/resource-auth-method/resource-auth-method-service.ts:983
before the PRs testedno
Turning Agent Vault session logs off can be silently undone by another save
s13-session-logs-turned-off-come-back-on-1
agent-vaultbehaviourloss: decision
medium

When Session logs are turned off while someone else saves a new key prefix at the same moment.

Got Both saves say they worked, but session logs are back on.

Want Session logs stay off with the new prefix, or one save is told to retry.

steps
  1. As an org admin, create an AWS app connection (LocalStack in the sandbox) and an S3 bucket.
  2. Turn Agent Vault session logs on with that connection, the bucket, the region and key prefix p1.
  3. Control: save {enabled: false}, then save {keyPrefix: "p2"} one after the other; the settings read back off with p2.
  4. Turn session logs back on with key prefix p1.
  5. Save {enabled: false} and, while it is in progress, save {keyPrefix: "p2"} (in the test the turn-off is held at its write by a row lock on the session log config, the prefix save fires while it is held, then both are released).
  6. Both saves answer 200. Read the session log settings back.
s13-session-logs-turned-off-come-back-on: turning session logs off and saving the key prefix together keeps both changes
e2e/specs/agent-vault/s13-session-logs-turned-off-come-back-on.spec.ts:21
lines
infisical/backend/src/ee/services/agent-vault-session-log/agent-vault-session-log-service.ts:511
infisical/backend/src/ee/services/agent-vault-session-log/agent-vault-session-log-service.ts:514
infisical/backend/src/ee/services/agent-vault-session-log/agent-vault-session-log-service.ts:593
infisical/backend/src/ee/routes/v1/agent-vault-routers/agent-vault-settings-router.ts:39
before the PRs testedno
Incident contacts look empty when they fail to load
s11-failed-incident-contacts-shown-as-empty-1
org-settingsbehaviourloss: decision
medium

When You open General settings and the incident contacts cannot be loaded.

Got The card says there are no incident contacts, as if none were ever added.

Want The card says the contacts could not be loaded.

steps
  1. Sign in as an organization admin.
  2. Have the incident contact list request fail with a server error.
  3. Open Organization Settings on the General tab and look at the Incident Contacts card.
a failed incident contacts load is not shown as an empty list
e2e/specs/org-settings/s11-failed-incident-contacts-shown-as-empty.spec.ts:15
lines
infisical/frontend/src/pages/organization/SettingsPage/components/OrgIncidentContactsSection/OrgIncidentContactsTable.tsx:42
infisical/frontend/src/pages/organization/SettingsPage/components/OrgIncidentContactsSection/OrgIncidentContactsTable.tsx:82
infisical/frontend/src/hooks/api/incidentContacts/queries.tsx:11
before the PRs testedno
One session with an unrecognised status blanks the whole Sessions page
s13-unknown-session-status-breaks-sessions-page-1
agent-vaultbehaviourloss: time
low

When One of your agent sessions has a status this version of the page does not recognise.

Got The Sessions page shows only "The page broke" and none of your sessions.

Want The page lists your sessions and shows the unfamiliar status plainly.

steps
  1. Sign in as an org admin of an organisation with Agent Vault.
  2. Create an access bundle and an agent session on it.
  3. Have the session list name that session's status as one the page does not know (here "suspended"; everything else unchanged).
  4. Open Agent Vault, Sessions.
s13-unknown-session-status-breaks-sessions-page
e2e/specs/agent-vault/s13-unknown-session-status-breaks-sessions-page.spec.ts:7
lines
infisical/frontend/src/pages/agent-vault/AgentVaultSessionsPage/components/SessionStatusBadge.tsx:28
infisical/frontend/src/pages/agent-vault/AgentVaultSessionsPage/AgentVaultSessionsPage.tsx:449
before the PRs testedno
Saving the organization name does nothing, with no message, when the role list failed to load
s1-org-name-save-silent-when-roles-fail-1
behaviourloss: time
medium

When

Got

Want

steps
  1. Sign in as an organization admin.
  2. Make the organization roles list request fail with a server error.
  3. Open Organization Settings, General tab; the Default Organization Member Role select stays on "Select Role" with no options.
  4. Change the Organization Name and click Save.
lines
infisical/frontend/src/pages/organization/SettingsPage/components/OrgNameChangeSection/OrgNameChangeSection.tsx:81
infisical/frontend/src/pages/organization/SettingsPage/components/OrgNameChangeSection/OrgNameChangeSection.tsx:64
before the PRs testedno
Two organizations can end up with the same slug when renamed at the same moment
s11-concurrent-org-renames-share-slug-1
org-settingsbehaviourloss: decision
low

When Two organizations are given the same slug at the same moment.

Got Both keep it, though giving a taken slug one after the other is refused.

Want One keeps the slug and the other is told it is already taken.

steps
  1. Sign in as an admin of two root organizations A and B (create B from the organization switcher).
  2. Rename A's slug to s on Organization Settings > General, then rename B's slug to s: the second is refused as 'Organization with slug s already exists'.
  3. Rename A and B to a fresh slug s2 at the same moment (the test holds both organization rows so both renames read before either writes).
  4. Count the root organizations whose slug is s2.
one slug is held by one root organization after two renames arrive together
e2e/specs/org-settings/s11-concurrent-org-renames-share-slug.spec.ts:12
lines
infisical/backend/src/services/org/org-service.ts:622
infisical/backend/src/services/org/org-service.ts:679
infisical/backend/src/db/migrations/20251018061215_sub-org.ts:18
infisical/backend/src/db/migrations/20251018061215_sub-org.ts:19
before the PRs testedno
A project template can keep a user on a role the template no longer has
s11-template-role-removed-while-user-given-it-1
org-settingsbehaviourloss: decision
low

When Two admins edit the same project template at once: one removes a role, the other gives a user that role.

Got Both saves succeed. The template lists a user with a role the template no longer has.

Want One of the two saves is refused, as it is when they happen one after the other.

steps
  1. As an org admin, create a project template with custom roles "racer" and "keeper" and no users.
  2. Send an edit that sets the template's roles to only "keeper".
  3. At the same moment, send an edit that adds yourself to the template's users with role "racer".
  4. Read the template's roles and its user memberships.
a template role removed while a user is given it leaves no user holding the removed role
e2e/specs/org-settings/s11-template-role-removed-while-user-given-it.spec.ts:23
lines
infisical/backend/src/ee/services/project-template/project-template-service.ts:543
infisical/backend/src/ee/services/project-template/project-template-service.ts:598
infisical/backend/src/ee/services/project-template/project-template-service.ts:735
infisical/backend/src/ee/services/project-template/project-template-service.ts:744
infisical/backend/src/ee/services/project-template/project-template-service.ts:764
before the PRs testedno
A product's project templates list says "No project templates found" when the templates fail to load
s11-failed-product-template-list-shown-as-empty-1
org-settingsbehaviourloss: decision
medium

When The project templates on a product's settings page fail to load.

Got The page says "No project templates found", as if the organization had none.

Want The page says the templates could not be loaded.

steps
  1. Sign in as an organization admin whose organization has project templates enabled.
  2. Create a KMS project template.
  3. Make the project template list request fail with a server error (GET /api/v1/project-templates answers 500).
  4. Open Organization Settings, Product Settings, then the KMS link (the KMS product settings page).
  5. Wait for the read to give up (it is retried once).
a failed project template load is not shown as an empty template list
e2e/specs/org-settings/s11-failed-product-template-list-shown-as-empty.spec.ts:18
lines
infisical/frontend/src/pages/organization/ProductSettingsPage/project-templates/ProjectTemplatesTable.tsx:49
infisical/frontend/src/pages/organization/ProductSettingsPage/project-templates/ProjectTemplatesTable.tsx:67
infisical/frontend/src/pages/organization/ProductSettingsPage/project-templates/ProjectTemplatesTable.tsx:86
infisical/frontend/src/hooks/api/projectTemplates/queries.tsx:30
before the PRs testedno
Double-clicking Create Session in the Agent Vault Create Session dialog creates two agent sessions
s3-double-clicked-create-session-makes-two-1
behaviourloss: time
low

When

Got

Want

steps
  1. Sign in as an org admin with Agent Vault and create an access bundle
  2. Open Agent Vault > Sessions and click Create Session
  3. Pick the access bundle in the dialog and keep the default expiry
  4. Double-click the dialog's Create Session button
  5. Search the Sessions list for the bundle's name
lines
infisical/frontend/src/pages/agent-vault/AgentVaultSessionsPage/components/CreateSessionDialog.tsx:112
infisical/frontend/src/pages/agent-vault/AgentVaultSessionsPage/components/CreateSessionDialog.tsx:241
infisical/backend/src/ee/routes/v1/agent-vault-routers/agent-vault-session-router.ts:124
before the PRs testedno
Pressing Escape in the Create Signer wizard closes it with no warning and discards the typed signer name
s9-create-signer-draft-discarded-on-esc-1
behaviourloss: time
low

When

Got

Want

steps
  1. Open a Certificate Manager project's Code Signing screen as the project admin
  2. Open Create Signer and type a signer name (placeholder e.g. mobile-app-prod) on the Basics step
  3. Press Escape
  4. Open Create Signer again: no confirmation was shown, and the name field is empty
lines
infisical/frontend/src/pages/cert-manager/CodeSigningPage/components/CreateSignerWizard/CreateSignerWizard.tsx:192
infisical/frontend/src/pages/cert-manager/CodeSigningPage/components/CreateSignerWizard/CreateSignerWizard.tsx:354
before the PRs testedno
A template role another admin just saved is erased by a role saved from an older page
s12-template-role-dropped-by-stale-role-save-1
org-settingsbehaviourloss: work
low

When Two admins (or two tabs) have the same project template open and each adds a role.

Got The second save quietly removes the role the first admin just added.

Want Both roles stay, or the second admin is told the template changed.

steps
  1. Create a project template with no custom roles.
  2. Open the template's page in two tabs.
  3. In tab 1, click Add Role, name it r1-first (slug r1-first) and Save.
  4. In tab 2, without reloading, click Add Role, name it r2-second (slug r2-second) and Save.
  5. Read the template's roles.
a template role saved by one admin survives a role saved from a page opened before it
e2e/specs/org-settings/s12-template-role-dropped-by-stale-role-save.spec.ts:20
lines
infisical/frontend/src/pages/organization/ProductSettingsPage/project-templates/EditProjectTemplateSection/components/ProjectTemplateEditRoleForm.tsx:84
infisical/frontend/src/pages/organization/ProductSettingsPage/project-templates/EditProjectTemplateSection/components/ProjectTemplateRolesSection.tsx:79
infisical/backend/src/ee/services/project-template/project-template-service.ts:744
before the PRs testedno
Closing a signer's Add member dialog throws away the picked identity
s14-add-member-pick-discarded-on-esc-1
code-signingbehaviourloss: time
low

When You pick members in a signer's Add member dialog and press Escape.

Got The dialog closes without asking, and when you reopen it your picks are gone.

Want It asks before discarding your picks, or keeps them.

steps
  1. Open a code signer in a Certificate Manager project as the project admin, on the Members tab
  2. Click Add Member, choose Machine Identities, and pick a machine identity of the project
  3. Press Escape (the first closes the picker list, the second closes the dialog)
  4. Click Add Member again and choose Machine Identities: no confirmation was shown, and the picker reads Pick... again
s14-add-member-pick-discarded-on-esc
e2e/specs/code-signing/s14-add-member-pick-discarded-on-esc.spec.ts:8
lines
infisical/frontend/src/pages/cert-manager/SignerDetailPage/components/AddSignerMemberModal.tsx:131
infisical/frontend/src/pages/cert-manager/SignerDetailPage/components/AddSignerMemberModal.tsx:123
before the PRs testedno
Deleting a gateway twice at once answers the second delete with a server error
s18-concurrent-gateway-delete-server-error-1
gateways-relaysbehaviourloss: time
low

When Two people delete the same gateway at the same moment.

Got The gateway is deleted, but one of them is told the delete failed.

Want The second person is told the gateway is already gone.

steps
  1. Sign in as an organisation admin and create a gateway.
  2. Send two deletes of that gateway so that both arrive before either finishes (the test holds the gateway's row so both requests reach the delete).
  3. Read the two answers.
s18-concurrent-gateway-delete-server-error
e2e/specs/gateways-relays/s18-concurrent-gateway-delete-server-error.spec.ts:9
lines
infisical/backend/src/ee/routes/v2/gateway-router.ts:216
infisical/backend/src/ee/services/gateway-v2/gateway-v2-service.ts:1120
infisical/backend/src/ee/services/gateway-v2/gateway-v2-service.ts:1147
infisical/backend/src/lib/knex/index.ts:302
before the PRs testedno
Pressing Escape in a signer's approval policy editor throws away the drafted policy
s15-approval-policy-draft-discarded-on-esc-1
code-signingbehaviourloss: time
low

When You draft an approval policy for a signer — a step, an approver, a signing limit — and press Escape.

Got The editor closes without asking. When you open it again, the step and the limit are gone and you start over.

Want The editor asks before discarding your draft, or keeps it.

steps
  1. Open a code signer that has no approval policy and go to its Approvals tab.
  2. Click Add Policy, add a step and pick an approver, click Continue.
  3. Type 7 in the max signings field.
  4. Press Escape.
  5. Click Add Policy again.
s15-approval-policy-draft-discarded-on-esc
e2e/specs/code-signing/s15-approval-policy-draft-discarded-on-esc.spec.ts:6
lines
infisical/frontend/src/pages/cert-manager/SignerDetailPage/components/EditSignerPolicyModal/EditSignerPolicyModal.tsx:140
infisical/frontend/src/pages/cert-manager/SignerDetailPage/components/EditSignerPolicyModal/EditSignerPolicyModal.tsx:79
infisical/frontend/src/pages/cert-manager/SignerDetailPage/components/EditSignerPolicyModal/EditSignerPolicyModal.tsx:229
before the PRs testedno
Pressing Escape in an HSM connector's Edit Details sheet discards the typed description
s15-hsm-connector-description-discarded-on-esc-1
code-signingbehaviourloss: time
low

When You type a new description for an HSM connector and press Escape.

Got The edit panel closes without asking, and your text is gone when you reopen it.

Want You are asked before your text is thrown away, or it is still there.

steps
  1. Open a Certificate Manager project's HSM connector page as the project admin
  2. Open Connector options, then Edit Details
  3. Type a description
  4. Press Escape
  5. Open Edit Details again: no confirmation was shown, and the description field is empty
s15-hsm-connector-description-discarded-on-esc
e2e/specs/code-signing/s15-hsm-connector-description-discarded-on-esc.spec.ts:38
lines
infisical/frontend/src/pages/cert-manager/SettingsPage/components/HsmConnectorsTab/EditHsmConnectorSheet.tsx:129
infisical/frontend/src/pages/cert-manager/SettingsPage/components/HsmConnectorsTab/EditHsmConnectorSheet.tsx:73
before the PRs testedno
A failed load of a PAM account's members shows "No members assigned to this account yet."
s3-failed-account-members-shown-as-no-members-1
behaviourloss: decision
high

When

Got

Want

steps
  1. Create a PAM folder and an AWS IAM account in it
  2. Assign yourself to the account directly as operator
  3. Open the account's detail sheet while the account's members fail to load (server error)
  4. Look at the Permissions tab
lines
infisical/frontend/src/pages/pam/PamAccountsPage/components/AccountDetailSheet.tsx:133
infisical/frontend/src/pages/pam/PamAccountsPage/components/AccountDetailSheet.tsx:216
infisical/frontend/src/pages/pam/PamAccountsPage/components/AccountDetailSheet.tsx:353
before the PRs testedno
Adding a user to a project template loads every member of the organization to check that one user
s13-template-user-add-reads-every-org-member-1
org-settingsperformanceloss: time
low

When You add a user to a project template in an organization with many members.

Got Infisical loads every member of the organization to check that one user. The work grows with every member added.

Want Saving the template looks up only the users you chose.

steps
  1. Sign in as an organization admin and create a project template.
  2. Add one user (the admin's own username) to the template, counting the rows of the organization-member query.
  3. Invite 45 more people to the organization.
  4. Add the same one user to the template again, counting the same query.
s13-template-user-add-reads-every-org-member
e2e/specs/org-settings/s13-template-user-add-reads-every-org-member.spec.ts:16
lines
infisical/backend/src/ee/services/project-template/project-template-service.ts:588
infisical/backend/src/ee/services/project-template/project-template-service.ts:306
infisical/backend/src/services/org-membership/org-membership-dal.ts:167
before the PRs testedno
Opening one project template downloads every template in the organization
s13-template-page-loads-every-template-1
org-settingsperformanceloss: time
medium

When You open one project template's page in an organization with many templates.

Got The page downloads every template in the organization, with all their members, before showing the one you opened.

Want The page loads only the template you opened.

steps
  1. Sign in as an organization admin whose organization has 3 project templates
  2. Open one template's page under Product Settings › Project Templates
  3. Note the template list the page fetches: every template of the organization, each with its roles, users, groups and identities
  4. Add 27 more templates and open the same template's page again
a project template's page reads only that template at 3 and 30 templates
e2e/specs/org-settings/s13-template-page-loads-every-template.spec.ts:26
lines
infisical/frontend/src/pages/organization/ProductSettingsPage/project-templates/EditProjectTemplateSection/EditProjectTemplateSection.tsx:15
infisical/frontend/src/pages/organization/ProductSettingsPage/project-templates/ProjectTemplatePage.tsx:30
infisical/frontend/src/pages/organization/ProductSettingsPage/project-templates/ProjectTemplatePage.tsx:29
before the PRs testedno
Turning cross-project secret sharing on or off gets slower with every project
s14-sharing-toggle-clears-cache-per-project-1
org-settingsperformanceloss: time
medium

When An admin turns cross-project secret sharing on or off for an organization.

Got The save does extra work for every project in the organization, so it gets slower and heavier as the organization grows.

Want The setting saves in the same time whatever the number of projects.

steps
  1. Sign in as an organization admin in an organization with 3 secret management projects (6 live projects in all, counting the defaults).
  2. Turn cross-project secret sharing on, then off, in the Secrets Management product settings, counting the database statements each toggle runs.
  3. Create 27 more secret management projects in the organization.
  4. Turn the setting on and off again, counting the statements once more.
lines
infisical/backend/src/services/org/org-service.ts:705
infisical/backend/src/services/org/org-service.ts:707
infisical/backend/src/services/secret-v2-bridge/secret-v2-bridge-dal.ts:56
infisical/backend/src/services/secret-v2-bridge/secret-v2-bridge-dal.ts:62
infisical/backend/src/services/project/project-dal.ts:594
before the PRs testedno
A PAM account that fails to load, or no longer exists, opens as a blank sheet with no message
s4-failed-account-load-sheet-blank-1
uiloss: time
medium

When

Got

Want

steps
  1. Create a PAM folder and an AWS IAM account in it
  2. Open the account's link: Accounts > the folder > the account (the URL carries ?accountId=)
  3. Make the account's details fail to load (the server answers an error), or delete the account first and open the old link
  4. Look at the account sheet
lines
infisical/frontend/src/pages/pam/PamAccountsPage/components/AccountDetailSheet.tsx:866
infisical/frontend/src/pages/pam/PamAccountsPage/components/AccountDetailSheet.tsx:930
infisical/frontend/src/pages/pam/PamAccountsPage/components/AccountDetailSheet.tsx:577
before the PRs testedno
A failed load of PAM discovery sources shows "No discovery sources yet" instead of an error
s5-failed-discovery-sources-shown-as-no-sources-1
uiloss: decision
medium

When

Got

Want

steps
  1. Sign in as a member of an organisation with PAM on the enterprise plan
  2. Make the load of the discovery source list fail (the server answers an error)
  3. Open PAM > Discovery
  4. Wait for the page to settle
lines
infisical/frontend/src/pages/pam/PamDiscoveryPage/PamDiscoveryPage.tsx:63
infisical/frontend/src/pages/pam/PamDiscoveryPage/PamDiscoveryPage.tsx:150
infisical/frontend/src/hooks/api/pam/queries.tsx:558
before the PRs testedno
A failed load of PAM account templates shows "No templates yet" instead of an error
s4-failed-template-list-shown-as-no-templates-1
behaviourloss: decision
medium

When

Got

Want

steps
  1. Sign in to an organisation with PAM on the enterprise plan that has at least one account template
  2. Make the account-templates list fail to load (a server error)
  3. Open PAM > Account Templates
lines
infisical/frontend/src/pages/pam/PamTemplatesPage/PamTemplatesPage.tsx:142
infisical/frontend/src/pages/pam/PamTemplatesPage/PamTemplatesPage.tsx:256
infisical/frontend/src/hooks/api/pam/queries.tsx:287
before the PRs testedno
The Add Account dialog says "No templates yet" when the account templates fail to load
s5-add-account-failed-templates-shown-as-no-templates-1
behaviourloss: decision
medium

When

Got

Want

steps
  1. Sign in to an organisation with PAM and at least one account template
  2. Make the account templates list fail to load (a server error)
  3. Open PAM > Accounts and click Add Account
  4. Look at the template picker in the dialog
lines
infisical/frontend/src/pages/pam/PamAccountsPage/components/CreateAccountSheet.tsx:172
infisical/frontend/src/pages/pam/PamAccountsPage/components/CreateAccountSheet.tsx:540
infisical/frontend/src/hooks/api/pam/queries.tsx:287
before the PRs testedno
Going Back in "Add Discovery Source" throws away the typed source name
s6-discovery-source-name-wiped-on-back-1
behaviourloss: time
low

When

Got

Want

steps
  1. Sign in as an organization admin with PAM available and open PAM > Discovery.
  2. Click Add Source, pick the first source type and click Next.
  3. Type a source name.
  4. Click Back.
  5. Click Next again with the same source type still picked.
lines
infisical/frontend/src/pages/pam/PamDiscoveryPage/components/AddDiscoverySourceSheet.tsx:427
infisical/frontend/src/pages/pam/PamDiscoveryPage/components/AddDiscoverySourceSheet.tsx:431
infisical/frontend/src/pages/pam/PamDiscoveryPage/components/AddDiscoverySourceSheet.tsx:243
infisical/frontend/src/pages/pam/PamDiscoveryPage/components/AddDiscoverySourceSheet.tsx:149
before the PRs testedno
A failed authenticator-app setup shows a blank QR code and a code box instead of an error
s1-totp-setup-load-failure-shows-blank-qr-1
behaviourloss: time
medium

When

Got

Want

steps
  1. Sign in as a user with no authenticator app configured
  2. Have the authenticator registration request fail on the server
  3. Open Personal settings, Authentication tab
  4. Click Set up on the Authenticator app row
  5. Wait for the registration request (and its one retry) to fail
lines
infisical/frontend/src/components/mfa/setup/VerifyStep.tsx:33
infisical/frontend/src/components/mfa/setup/VerifyStep.tsx:61
infisical/frontend/src/components/mfa/setup/VerifyStep.tsx:74
infisical/frontend/src/hooks/api/users/queries.tsx:490
infisical/frontend/src/pages/user/PersonalSettingsPage/components/SecuritySection/MethodSetupDialog.tsx:31
before the PRs testedno
Double-clicking Add in a gateway pool adds the gateway but shows a failure message
s8-double-clicked-pool-add-reports-failure-1
behaviourloss: time
low

When

Got

Want

steps
  1. Sign in as an organisation admin and open Networking > Gateway Pools
  2. Create a gateway pool and open its sheet
  3. Pick a connected gateway in Member Gateways
  4. Double-click Add
lines
infisical/frontend/src/pages/organization/NetworkingPage/components/GatewayTab/components/PoolDetailSheet.tsx:91
infisical/frontend/src/pages/organization/NetworkingPage/components/GatewayTab/components/PoolDetailSheet.tsx:196
infisical/backend/src/ee/services/gateway-pool/gateway-pool-service.ts:239
before the PRs testedno
Double-clicking Delete Relay deletes the relay but shows "Failed to delete relay"
s7-double-clicked-relay-delete-reports-failure-1
behaviourloss: time
medium

When

Got

Want

steps
  1. Sign in as an organisation admin and open Organization Settings > Networking > Relays.
  2. Create a relay and open its detail page.
  3. Open Options > Delete Relay and type the relay's name.
  4. Double-click the Delete Relay confirm button.
  5. The page goes back to Networking and the relay is gone, but a "Failed to delete relay" notification is shown; the second delete request answered a server error.
lines
infisical/frontend/src/pages/organization/NetworkingPage/RelayDetailsByIDPage/components/RelayPageHeader/RelayPageHeader.tsx:152
infisical/frontend/src/pages/organization/NetworkingPage/RelayDetailsByIDPage/components/RelayPageHeader/RelayPageHeader.tsx:58
infisical/backend/src/ee/services/relay/relay-service.ts:1218
infisical/backend/src/ee/routes/v1/relay-router.ts:165
before the PRs testedno
Double-clicking Delete Gateway on a gateway's page deletes it but also says the delete failed
s19-double-clicked-gateway-delete-shows-failure-1
gateways-relaysbehaviourloss: nothing
low

When You double-click Delete Gateway on a gateway's page.

Got The gateway is deleted and you are told so, but you are also told the delete failed.

Want Only the success message, since the gateway was deleted.

steps
  1. Sign in as an organisation admin and create a gateway from Networking.
  2. On the gateway's detail page open Options, choose Delete Gateway and type the gateway's name.
  3. Double-click the Delete Gateway confirm button.
  4. Look at the notifications once the page is back on Networking.
s19-double-clicked-gateway-delete-shows-failure
e2e/specs/gateways-relays/s19-double-clicked-gateway-delete-shows-failure.spec.ts:5
lines
infisical/frontend/src/pages/organization/NetworkingPage/GatewayDetailsByIDPage/components/GatewayPageHeader/GatewayPageHeader.tsx:199
infisical/frontend/src/pages/organization/NetworkingPage/GatewayDetailsByIDPage/components/GatewayPageHeader/GatewayPageHeader.tsx:64
before the PRs testedno
Deleting a gateway makes the app fetch the gateway it just deleted
s19-deleted-gateway-requested-again-1
gateways-relaysbehaviourloss: nothing
low

When You delete a gateway from its detail page.

Got The gateway is deleted and you land on the list, but the app quietly asks for the deleted gateway again and logs errors.

Want The app stops asking for a gateway once it is deleted.

steps
  1. Sign in as an organisation admin and open Networking.
  2. Register a gateway and open its detail page.
  3. Open the gateway's options, choose Delete Gateway, type its name and confirm.
  4. Watch the requests the page sends after the delete is answered.
a gateway deleted from its detail page is not requested again
e2e/specs/gateways-relays/s19-deleted-gateway-requested-again.spec.ts:5
lines
infisical/frontend/src/hooks/api/gateways-v2/mutations.tsx:19
infisical/frontend/src/hooks/api/gateways-v2/mutations.tsx:21
infisical/frontend/src/hooks/api/gateways-v2/mutations.tsx:34
infisical/frontend/src/pages/organization/NetworkingPage/GatewayDetailsByIDPage/components/GatewayPageHeader/GatewayPageHeader.tsx:60
before the PRs testedno
The gateway list runs four database counts for every gateway
s17-gateway-list-counts-per-gateway-1
gateways-relaysperformanceloss: time
medium

When You open Networking in an organisation with many gateways.

Got Loading the gateway list does four separate lookups for every gateway, so it gets slower with each gateway you add.

Want The list looks up what is connected to all gateways at once.

steps
  1. Sign in as an organisation admin in a fresh organisation
  2. Create thirty gateways with token authentication
  3. List the organisation's gateways once, counting the database statements the list runs
s17 listing an organisation's 30 gateways counts connected resources for all gateways together
e2e/specs/gateways-relays/s17-gateway-list-counts-per-gateway.spec.ts:26
lines
infisical/backend/src/ee/services/gateway-v2/gateway-v2-service.ts:334
infisical/backend/src/ee/services/gateway-v2/gateway-v2-service.ts:339
infisical/backend/src/services/app-connection/app-connection-dal.ts:173
infisical/backend/src/ee/services/dynamic-secret/dynamic-secret-dal.ts:245
infisical/backend/src/services/identity-kubernetes-auth/identity-kubernetes-auth-dal.ts:38
infisical/backend/src/ee/services/pki-discovery/pki-discovery-config-dal.ts:169
before the PRs testedno
A gateway switched to AWS login can still enroll with a token made during the switch
s20-token-generated-during-aws-switch-still-enrolls-1
gateways-relaysbehaviourloss: decision
medium

When You switch a gateway to AWS login while a colleague generates an enrollment token for it.

Got Both succeed. The gateway shows AWS login, yet the new token still enrolls it.

Want The token is refused or removed, and only AWS logins enroll the gateway.

steps
  1. Create a gateway on token authentication in an organisation.
  2. Start generating an enrollment token for it, and while that request is between its method check and its token write, switch the gateway to AWS authentication.
  3. Both requests succeed.
  4. Log in as a gateway with the enrollment token that was returned.
an enrollment token generated during a switch to AWS authentication cannot enroll the gateway
e2e/specs/gateways-relays/s20-token-generated-during-aws-switch-still-enrolls.spec.ts:13
lines
infisical/backend/src/ee/services/resource-auth-method/resource-auth-method-service.ts:1078
infisical/backend/src/ee/services/resource-auth-method/resource-auth-method-service.ts:1089
infisical/backend/src/ee/services/resource-auth-method/resource-auth-method-service.ts:983
infisical/backend/src/ee/services/resource-auth-method/resource-auth-method-service.ts:1393
before the PRs testedno
The gateway pool list runs four database counts for every pool
s11-pool-list-counts-resources-per-pool-1
performanceloss: time
medium

When

Got

Want

steps
  1. Sign in as an organisation admin with gateway pools enabled
  2. Create eighteen or more gateway pools
  3. List the organisation's pools twice, counting the database statements
  4. Each of the four resource-count statements has run once per pool per list
lines
infisical/backend/src/ee/services/gateway-pool/gateway-pool-service.ts:119
infisical/backend/src/services/app-connection/app-connection-dal.ts:197
infisical/backend/src/services/identity-kubernetes-auth/identity-kubernetes-auth-dal.ts:60
infisical/backend/src/ee/services/pki-discovery/pki-discovery-config-dal.ts:192
infisical/backend/src/ee/services/dynamic-secret/dynamic-secret-dal.ts:274
before the PRs testedno
The admin Organizations list skips root organizations when sub-organizations exist, so pages come back short and some organizations are never listed
s1-admin-org-list-pages-skip-orgs-1
behaviourloss: decision
medium

When

Got

Want

steps
  1. As the super admin, create a sub-organization whose name sorts before a root organization's name, both sharing one search term
  2. Create the root organization from the Server Console's Organizations tab
  3. List the Organizations tab for that search term with a large page: one organization, total 1
  4. Page the same list one organization at a time for every offset below the total: the page at offset 0 is empty and the root organization never appears
lines
infisical/backend/src/services/org/org-dal.ts:60
infisical/backend/src/services/org/org-dal.ts:73
infisical/backend/src/services/org/org-dal.ts:78
infisical/backend/src/services/org/org-dal.ts:109
infisical/frontend/src/pages/admin/ResourceOverviewPage/components/OrganizationsTable.tsx:471
before the PRs testedno
A failed list load on the Server Console's Resource Overview shows "No users found" (or no organizations, identities, email domains) instead of an error
s2-admin-resource-list-failure-shown-as-empty-1
behaviourloss: decision
medium

When

Got

Want

steps
  1. Sign in as the instance super admin on an instance whose setup is finished
  2. Make the list request behind a Resource Overview tab fail with a server error (Users: /api/v1/admin/user-management/users; also Organizations, Machine Identities, Email Domains)
  3. Open the Server Console's Resource Overview on that tab
  4. Wait for the load and its single retry to fail
lines
infisical/frontend/src/pages/admin/ResourceOverviewPage/components/UserIdentitiesTable.tsx:358
infisical/frontend/src/pages/admin/ResourceOverviewPage/components/UserIdentitiesTable.tsx:112
infisical/frontend/src/pages/admin/ResourceOverviewPage/components/UserIdentitiesTable.tsx:311
infisical/frontend/src/pages/admin/ResourceOverviewPage/components/OrganizationsTable.tsx:479
infisical/frontend/src/pages/admin/ResourceOverviewPage/components/MachineIdentitiesTable.tsx:81
infisical/frontend/src/pages/admin/ResourceOverviewPage/components/EmailDomainsTable.tsx:216
before the PRs testedno
Expired sessions stay listed as having access to your account
s7-expired-session-listed-as-active-1
personal-settingsbehaviourloss: decision
low

When One of your sign-ins expires because it went unused longer than your organisation's session length.

Got Personal settings still lists it among the sessions with access to your account, with a Sign out button.

Want Only sessions that can still be used are listed, or expired ones are marked as expired.

steps
  1. Sign up a fresh account with its own organisation.
  2. In Organization Settings, Security, set the session length to 1 minute.
  3. Sign in a second time from another browser or client.
  4. Wait more than one minute without using that second sign-in, so it expires and can no longer be renewed.
  5. Sign in again and open Personal settings, General tab, and read the Sessions table.
s7-expired-session-listed-as-active
e2e/specs/personal-settings/s7-expired-session-listed-as-active.spec.ts:12
lines
infisical/backend/src/services/auth-token/auth-token-dal.ts:70
infisical/backend/src/services/auth-token/auth-token-service.ts:252
infisical/backend/src/server/routes/v2/user-router.ts:416
infisical/frontend/src/pages/user/PersonalSettingsPage/components/SessionsSection/SessionsTable.tsx:130
before the PRs testedno
The admin Organizations list reads every project paired with every member
s8-admin-org-list-reads-projects-times-members-1
server-adminperformanceloss: time
medium

When You open the Organizations tab of the Server Console on an instance whose organizations have many projects and members.

Got The list gets slower much faster than the organizations grow, because each organization's projects are read once per member.

Want The list costs about one read per project and per member.

steps
  1. Sign in as the instance super admin
  2. Create an organization in the admin console inviting ten members, then create ten projects in it
  3. Open Server Console > Resources > Organizations and search for that organization
  4. Read the rows the database returned for the list statement (pg_stat_statements)
s8-admin-org-list-reads-projects-times-members rows read stay within projects plus members
e2e/specs/server-admin/s8-admin-org-list-reads-projects-times-members.spec.ts:13
lines
infisical/backend/src/services/org/org-dal.ts:80
infisical/backend/src/services/org/org-dal.ts:81
infisical/backend/src/services/org/org-dal.ts:112
infisical/backend/src/server/routes/v1/admin-router.ts:261
infisical/frontend/src/pages/admin/ResourceOverviewPage/components/OrganizationsTable.tsx:562
before the PRs testedno
The Caching page says the cache was invalidated when it cannot tell
s9-caching-success-announced-on-failed-status-read-1
server-adminbehaviourloss: decision
medium

When You invalidate the secrets cache and the page cannot check on its progress.

Got The page tells you the cache was successfully invalidated, though it never learned whether it finished.

Want The page keeps showing the invalidation as running, or says it cannot check on it.

steps
  1. Sign in as the instance super admin and open the Server Console's Caching page.
  2. Have the invalidation status read fail with a server error.
  3. Click Invalidate secrets cache, type confirm and click Invalidate cache.
s9-caching-success-announced-on-failed-status-read no completion announced without a status
e2e/specs/server-admin/s9-caching-success-announced-on-failed-status-read.spec.ts:8
lines
infisical/frontend/src/pages/admin/CachingPage/components/CachingPageForm.tsx:47
infisical/frontend/src/pages/admin/CachingPage/components/CachingPageForm.tsx:76
infisical/frontend/src/hooks/api/admin/queries.ts:186
before the PRs testedno
A failed load of the encryption strategies leaves the KMS Encryption Strategy card with no strategy picker and no error
s2-encryption-strategy-load-failure-silent-1
uiloss: time
medium

When

Got

Want

steps
  1. Sign in as the super admin of a set-up instance
  2. Make the server fail to return the encryption strategies (here: the request answered with a server error)
  3. Open the admin console's Encryption page
  4. Look at the KMS Encryption Strategy card
lines
infisical/frontend/src/pages/admin/EncryptionPage/components/EncryptionPageForm.tsx:49
infisical/frontend/src/pages/admin/EncryptionPage/components/EncryptionPageForm.tsx:99
infisical/frontend/src/hooks/api/admin/queries.ts:142
before the PRs testedno
Turning on two-factor in two tabs shows recovery codes that never work
s8-two-tab-mfa-activation-shows-dead-recovery-codes-1
personal-settingsbehaviourloss: decision
medium

When You turn on two-factor from two tabs at about the same time.

Got Both tabs say it worked and each shows ten recovery codes. Only one tab's codes work; the other ten are dead, with no warning.

Want One tab turns it on and shows working codes; the other says two-factor is already on.

steps
  1. Sign up a fresh account; leave two-factor off.
  2. Open Personal settings > Authentication in two tabs and start turning on email two-factor in each; confirm the emailed step-up code.
  3. Submit the activation in both tabs at the same moment (the test holds the account's row lock so both arrive before either commits).
  4. Note the recovery codes each tab shows.
  5. Open the account's recovery codes from Personal settings (step-up confirmed).
s8-two-tab-mfa-activation-shows-dead-recovery-codes
e2e/specs/personal-settings/s8-two-tab-mfa-activation-shows-dead-recovery-codes.spec.ts:89
lines
infisical/backend/src/services/user/user-service.ts:211
infisical/backend/src/services/user/user-service.ts:216
infisical/backend/src/services/mfa-recovery-code/mfa-recovery-code-service.ts:110
infisical/backend/src/server/routes/v2/user-router.ts:199
before the PRs testedno
A failed load leaves the Overrides card empty with no error
s10-env-overrides-load-failure-shown-as-empty-1
server-adminuiloss: decision
medium

When You open Environment Variables in the admin console and the variables can't be loaded.

Got The Overrides card is just an empty search box and a Save button, as if there were nothing to override.

Want The card tells you the variables could not be loaded.

steps
  1. Sign in as the instance super admin on an instance whose setup is finished
  2. Open the Server Admin Console and go to Environment Variables while the server fails to return the overridable variables
  3. Look at the Overrides card
s10-env-overrides-load-failure-shown-as-empty the overrides load failure is reported
e2e/specs/server-admin/s10-env-overrides-load-failure-shown-as-empty.spec.ts:8
lines
infisical/frontend/src/pages/admin/EnvironmentPage/components/EnvironmentPageForm.tsx:103
infisical/frontend/src/pages/admin/EnvironmentPage/components/EnvironmentPageForm.tsx:216
infisical/frontend/src/hooks/api/admin/queries.ts:205
before the PRs testedno
Two-factor can stay enabled with no authenticator left to satisfy it
s8-enable-totp-while-removing-it-leaves-mfa-without-a-factor-1
personal-settingsbehaviourloss: work
medium

When You turn on two-factor with your authenticator app at the same moment you remove that authenticator.

Got Both go through. Your account now says two-factor is on with the authenticator as the method, but no authenticator is left.

Want One of the two is refused, so your account always keeps a usable factor.

steps
  1. Create a new account with a verified authenticator app and two-factor authentication off
  2. Prove the authenticator app to obtain a fresh step-up verification for enabling two-factor
  3. Fire the enable-two-factor request (authenticator method) and, while it is still running, fire the remove-authenticator request
  4. Read the account's two-factor state: two-factor is on, the authenticator is the selected method, and no authenticator remains configured
enabling the authenticator while it is being removed never leaves two-factor on with no factor
e2e/specs/personal-settings/s8-enable-totp-while-removing-it-leaves-mfa-without-a-factor.spec.ts:93
lines
infisical/backend/src/services/user/user-service.ts:214
infisical/backend/src/services/totp/totp-service.ts:197
infisical/backend/src/services/totp/totp-service.ts:199
infisical/backend/src/services/user/user-service.ts:189
before the PRs testedno
Creating two organizations at once for the same new admin loses one
s11-concurrent-org-creates-same-new-admin-lose-one-1
server-adminbehaviourloss: time
low

When Two organizations inviting the same new admin email are created from the admin console at the same moment.

Got One is created; the other fails with "Something went wrong" and never appears.

Want Both organizations are created with that person invited to each.

steps
  1. Sign in as an instance super admin and get an API token.
  2. Create two organizations one after the other from the admin console route, each inviting the same brand-new email as admin: both are created and the email is invited to both (the control).
  3. With another brand-new email, send two create-organization requests at the same moment, each inviting that email as admin (the test forces both to look the account up before either inserts it by holding writes to the users table).
  4. List the admin Organizations table searched to the two names.
s11-concurrent-org-creates-same-new-admin-lose-one
e2e/specs/server-admin/s11-concurrent-org-creates-same-new-admin-lose-one.spec.ts:9
lines
infisical/backend/src/services/super-admin/super-admin-service.ts:945
infisical/backend/src/services/super-admin/super-admin-service.ts:948
infisical/backend/src/services/super-admin/super-admin-service.ts:929
infisical/backend/src/db/migrations/20240311210135_ldap-config.ts:48
before the PRs testedno
A refused EnvKey import is announced as "Import started" and the dialog closes with no error
s1-refused-envkey-import-announced-as-started-1
behaviourloss: decision
medium

When

Got

Want

steps
  1. Sign in as an organisation admin and open Organization > Integrations > External Migrations
  2. Click Import and choose EnvKey
  3. Drop an EnvKey export file (here {"nonce":"x","data":"y"}) and type a decryption key that does not match it
  4. Click Import data
  5. The server refuses the import, but an "Import started" notice says an email will follow, and the dialog closes
lines
infisical/frontend/src/hooks/api/migration/mutations.tsx:39
infisical/frontend/src/hooks/api/migration/mutations.tsx:24
infisical/frontend/src/pages/organization/SettingsPage/components/ExternalMigrationsTab/components/EnvKeyPlatformModal.tsx:49
infisical/backend/src/services/external-migration/external-migration-service.ts:255
before the PRs testedno
Removing your two passkeys at once leaves passkey required with none registered
s9-last-passkeys-removed-together-keep-passkey-required-1
personal-settingsbehaviourloss: work
medium

When You remove both of your passkeys at about the same time (two tabs, or a retry).

Got Both say they worked and both passkeys are gone, but your account still requires a passkey to sign in — and you have none.

Want Removing your last passkey switches your two-factor method to email, or the second removal is refused.

steps
  1. Sign up a fresh account and register two passkeys in Personal settings, Authentication.
  2. Turn on two-factor and set passkey as the preferred (required) method.
  3. Remove both passkeys at the same moment (the test holds a row lock on the first passkey so both removals read the credential set before either commits).
  4. Read the account's selected two-factor method and passkey list.
s9-last-passkeys-removed-together-keep-passkey-required
e2e/specs/personal-settings/s9-last-passkeys-removed-together-keep-passkey-required.spec.ts:125
lines
infisical/backend/src/services/webauthn/webauthn-service.ts:357
infisical/backend/src/services/webauthn/webauthn-service.ts:358
infisical/backend/src/services/webauthn/webauthn-service.ts:361
infisical/backend/src/server/routes/v1/user-router.ts:591
infisical/backend/src/server/routes/v1/user-router.ts:598
before the PRs testedno
A discovery jobs search with no matches says the project has no jobs
s10-discovery-no-match-search-shown-as-no-jobs-1
uiloss: decision
low

When

Got

Want

steps
  1. Sign in as a member of a Certificate Manager project that has one certificate discovery job.
  2. Open the project's Discovery screen and wait for the job to be listed on the Discovery Jobs tab.
  3. Type a name that no job has into "Search by name, domain, or IP…".
  4. Look at the tab once the filtered list has loaded.
lines
infisical/frontend/src/pages/cert-manager/DiscoveryPage/components/DiscoveryJobsTab.tsx:217
infisical/frontend/src/pages/cert-manager/DiscoveryPage/components/DiscoveryJobsTab.tsx:220
before the PRs testedno
After a search narrows the Findings list, Update Status changes selected findings the search hid
s7-findings-bulk-update-includes-hidden-1
behaviourloss: work
medium

When

Got

Want

steps
  1. Open a Secret Scanning project's Findings page with several findings
  2. Tick the select-all checkbox in the table header
  3. Type github-pat into the Search findings box: one finding stays visible, still ticked
  4. Open Update Status from the selection bar, pick a status and save
lines
infisical/frontend/src/pages/secret-scanning/SecretScanningFindingsPage/components/SecretScanningFindingsTable.tsx:443
infisical/frontend/src/pages/secret-scanning/SecretScanningFindingsPage/components/SecretScanningFindingsTable.tsx:430
infisical/frontend/src/pages/secret-scanning/SecretScanningFindingsPage/components/SecretScanningFindingsTable.tsx:96
before the PRs testedno
An EnvKey import of a damaged export file fails with "Something went wrong" instead of saying the file is invalid
s1-malformed-envkey-file-crashes-import-1
behaviourloss: time
medium

When

Got

Want

steps
  1. Sign in as an organisation admin and open Organization > Integrations > External Migrations
  2. Click Import and choose EnvKey
  3. Drop a JSON file whose nonce and data are not base64, for example {"nonce":"not base64!","data":"also not base64!"}, and type any decryption key
  4. Click Import data
  5. The server answers 500 "Something went wrong" and the API log shows TypeError: invalid encoding
lines
infisical/backend/src/services/external-migration/external-migration-fns/envkey.ts:52
infisical/backend/src/server/routes/v3/external-migration-router.ts:50
before the PRs testedno
Cancelling Create Discovery Job throws away the typed job without asking
s9-discovery-job-cancel-discards-draft-1
uiloss: time
low

When

Got

Want

steps
  1. Sign in and open a Certificate Manager project's Discovery screen
  2. Click Add Job
  3. Type a job name, targets and ports
  4. Click Cancel
  5. Click Add Job again
lines
infisical/frontend/src/pages/cert-manager/DiscoveryPage/components/DiscoveryJobModal.tsx:500
infisical/frontend/src/pages/cert-manager/DiscoveryPage/components/DiscoveryJobModal.tsx:333
infisical/frontend/src/pages/cert-manager/DiscoveryPage/components/DiscoveryJobModal.tsx:267
before the PRs testedno
Going Back in Edit Data Source throws away a typed description without warning
s15-data-source-description-discarded-on-back-1
secret-scanningbehaviourloss: time
low

When You type a new description for a data source and click Back until the edit window closes.

Got It closes without asking, and your text is gone when you reopen it.

Want You are asked before your text is thrown away, or it is still there.

steps
  1. Open a Secret Scanning project's GitHub data source page as a project member
  2. Click Edit data source, then Next to reach the Details step
  3. Type a new description
  4. Click Back (the Configuration step shows), then Back again (the dialog closes)
  5. Click Edit data source and Next again
s15-data-source-description-discarded-on-back
e2e/specs/secret-scanning/s15-data-source-description-discarded-on-back.spec.ts:25
lines
infisical/frontend/src/components/secret-scanning/forms/SecretScanningDataSourceForm.tsx:86
infisical/frontend/src/components/secret-scanning/EditSecretScanningDataSourceModal.tsx:36
before the PRs testedno
Unsaved scanning configuration is lost without warning on leaving Settings
s15-unsaved-scanning-config-lost-leaving-to-org-settings-1
secret-scanningbehaviourloss: time
low

When You edit the scanning configuration and leave Settings before saving.

Got Nothing warns you. When you come back, your edit is gone.

Want A warning before leaving, or your edit still there.

steps
  1. Sign in as an admin of a Secret Scanning project
  2. Open the project's Settings on the Scanning Settings tab
  3. Type a new line into the Project Configuration editor without saving
  4. Click Org settings in the sidebar
  5. Press the browser's Back button
an unsaved scanning configuration is not lost on leaving to Org settings
e2e/specs/secret-scanning/s15-unsaved-scanning-config-lost-leaving-to-org-settings.spec.ts:7
lines
infisical/frontend/src/pages/secret-scanning/SettingsPage/components/ProjectScanningConfigTab/SecretScanningConfigForm.tsx:33
before the PRs testedno
A honey token's details drawer says "No credentials available." or "Could not find honey token." when its data fails to load
s3-failed-honey-token-credentials-shown-as-none-1
uiloss: decision
medium

When

Got

Want

steps
  1. Plant an AWS honey token in a secrets project of your organisation.
  2. Make the honey token's credentials request fail with a server error (it and the app's retry).
  3. Open the project overview with the honey token's details drawer open (overview?honeyTokenId=<id>).
  4. The Credentials section reads "No credentials available." — no error.
  5. Instead make the honey token's own request fail with a server error and open the drawer again.
  6. The drawer reads "Could not find honey token." for a token that exists.
lines
infisical/frontend/src/components/honey-tokens/HoneyTokenDetailsDrawer.tsx:90
infisical/frontend/src/components/honey-tokens/HoneyTokenDetailsDrawer.tsx:270
infisical/frontend/src/components/honey-tokens/HoneyTokenDetailsDrawer.tsx:70
infisical/frontend/src/components/honey-tokens/HoneyTokenDetailsDrawer.tsx:115
before the PRs testedno
A project's Reports card shows "No reports yet" when its report list fails to load, hiding the reports that exist
s1-failed-report-list-shown-as-no-reports-1
behaviourloss: decision
medium

When

Got

Want

steps
  1. Open a secrets project that already has a report generated (Insights page, Reports card lists it)
  2. Make the project's report list fail to load with a server error (first try and the app's retry)
  3. Open the project's Insights page
  4. Look at the Reports card
lines
infisical/frontend/src/pages/secret-manager/InsightsPage/components/AuditReportsCard.tsx:103
infisical/frontend/src/pages/secret-manager/InsightsPage/components/AuditReportsCard.tsx:108
infisical/frontend/src/pages/secret-manager/InsightsPage/components/AuditReportsCard.tsx:158
before the PRs testedno
A project's Insights page that fails to load its summary reports "No failed rotations", "No overdue reminders" and "All secrets up to date"
s1-failed-insights-summary-shown-as-healthy-1
behaviourloss: decision
high

When

Got

Want

steps
  1. Sign in as a member of a secrets project
  2. Make the project's insights summary request fail with a server error, on the first try and the client's retry
  3. Open the project's Insights page
  4. The Upcoming Rotations, Upcoming Reminders and Stale Secrets cards show 0 with the green "No failed rotations", "No overdue reminders" and "All secrets up to date" badges, and no error
lines
infisical/frontend/src/pages/secret-manager/InsightsPage/components/InsightsSummaryCards.tsx:387
infisical/frontend/src/pages/secret-manager/InsightsPage/components/InsightsSummaryCards.tsx:406
infisical/frontend/src/pages/secret-manager/InsightsPage/components/InsightsSummaryCards.tsx:423
infisical/frontend/src/pages/secret-manager/InsightsPage/components/InsightsSummaryCards.tsx:467
before the PRs testedno
A project's Duplicated Secrets card says "No duplicated secrets found" when the duplicate check fails to load
s4-failed-duplicate-check-shown-as-all-unique-1
behaviourloss: decision
high

When

Got

Want

steps
  1. Create a secrets project with two secrets that share one value (DUP_A and DUP_B).
  2. Make the project's duplicate-secrets check fail with a server error (both the load and the app's automatic retry).
  3. Open the project's Insights page and look at the Duplicated Secrets card.
lines
infisical/frontend/src/pages/secret-manager/InsightsPage/components/DuplicatedSecretsCard.tsx:64
infisical/frontend/src/pages/secret-manager/InsightsPage/components/DuplicatedSecretsCard.tsx:67
infisical/frontend/src/pages/secret-manager/InsightsPage/components/DuplicatedSecretsCard.tsx:192
before the PRs testedno
Cancelling Create Access Bundle throws away the typed bundle name without asking
s4-bundle-name-draft-discarded-on-cancel-1
uiloss: time
low

When

Got

Want

steps
  1. Sign in as an organization admin and open Agent Vault, Access Bundles
  2. Click Create Access Bundle and type a bundle name
  3. Press Cancel
  4. Click Create Access Bundle again
lines
infisical/frontend/src/components/agent-vault/AccessBundleFormDialog.tsx:132
infisical/frontend/src/components/agent-vault/AccessBundleFormDialog.tsx:57
before the PRs testedno
A large EnvKey import stops anyone in the org from creating a project until it ends
s3-envkey-import-blocks-project-creation-1
external-migrationperformanceloss: time
medium

When Someone in the organisation is importing a large EnvKey export.

Got Anyone else who creates a project in that organisation is stuck waiting until the whole import finishes, and may see the request time out.

Want Creating a project works normally while an import runs.

steps
  1. Sign in as an org admin and start an EnvKey import of an export holding one project with 20,000 secrets.
  2. While the import is still running, create a new secret-manager project in the same organisation.
  3. Watch the database locks for the organisation's project-creation lock while the create is pending.
A project created while an EnvKey import runs is not held behind the import
e2e/specs/external-migration/s3-envkey-import-blocks-project-creation.spec.ts:52
lines
infisical/backend/src/services/external-migration/external-migration-fns/import.ts:52
infisical/backend/src/services/external-migration/external-migration-fns/import.ts:54
infisical/backend/src/services/project/project-service.ts:299
infisical/backend/src/services/project/project-service.ts:300
infisical/backend/src/services/external-migration/external-migration-fns/import.ts:346
before the PRs testedno
The org duplicated-values insight sends back every duplicated secret, with no cap
s8-org-duplicates-insight-has-no-cap-1
secret-insightsperformanceloss: time
medium

When Your organization has many secrets that share the same value.

Got The duplicated-values insight loads every one of them at once, so it gets slower and heavier with every copy.

Want It shows a bounded number at a time, like the search by value does.

steps
  1. Sign in as an org admin and create a new organization (org-wide secret value tracking is on from the start).
  2. Create two secrets projects and add 1100 secrets to each, all with the same value.
  3. Open the organization's Secret Insights and read the duplicated secret values insight (GET /api/v1/insights/secrets/secrets-duplication).
the org-wide duplicated-values insight returns at most 1000 secrets
e2e/specs/secret-insights/s8-org-duplicates-insight-has-no-cap.spec.ts:12
lines
infisical/backend/src/services/secret-v2-bridge/secret-v2-bridge-dal.ts:1596
infisical/backend/src/ee/services/insights/insights-service.ts:535
infisical/backend/src/ee/services/insights/insights-service.ts:156
before the PRs testedno
On the organisation's Insights page, a card whose data fails to load disappears with no error: the project health list, the summary cards and the New Secrets Created chart
s4-failed-org-projects-card-vanishes-1
uiloss: decision
medium

When

Got

Want

steps
  1. Sign in as an organisation admin with a secrets project.
  2. Make the organisation's project insights list fail to load with a server error (its first try and the app's retry).
  3. Open the organisation's Insights page (Secrets Management → Insights).
  4. The Needs Attention / All Projects card is missing: the summary cards are followed directly by Duplicated Secrets, and no error is shown.
  5. Do the same with the insights summary request: the Dynamic Secrets / Users / Machine Identities cards are missing, no error.
  6. Do the same with the static-secrets usage request: the New Secrets Created chart is missing, no error.
lines
infisical/frontend/src/pages/organization/SecretInsightsPage/SecretInsightsPage.tsx:89
infisical/frontend/src/pages/organization/SecretInsightsPage/SecretInsightsPage.tsx:247
infisical/frontend/src/pages/organization/SecretInsightsPage/SecretInsightsPage.tsx:86
infisical/frontend/src/pages/organization/SecretInsightsPage/SecretInsightsPage.tsx:245
infisical/frontend/src/pages/organization/SecretInsightsPage/SecretInsightsPage.tsx:121
infisical/frontend/src/pages/organization/SecretInsightsPage/SecretInsightsPage.tsx:279
before the PRs testedno
Cancelling Create Gateway throws away the typed name without asking
s21-create-gateway-name-discarded-on-cancel-1
gateways-relaysuiloss: time
low

When You start creating a gateway, type its name, then press Cancel.

Got The dialog closes straight away and the name you typed is gone when you open it again.

Want You are asked whether to discard what you typed, or the name is kept.

steps
  1. Sign in as an organisation admin and open Networking, Gateways tab
  2. Click Create Gateway and type a gateway name
  3. Press Cancel
  4. Click Create Gateway again
s21-create-gateway-name-discarded-on-cancel
e2e/specs/gateways-relays/s21-create-gateway-name-discarded-on-cancel.spec.ts:4
lines
infisical/frontend/src/pages/organization/NetworkingPage/components/GatewayTab/components/GatewayDeployModal.tsx:102
infisical/frontend/src/pages/organization/NetworkingPage/components/GatewayTab/components/GatewayDeployModal.tsx:122
before the PRs testedno
A project's duplicated-secrets check loads and returns every duplicated secret, with no cap
s8-project-duplicate-check-returns-every-secret-1
secret-insightsperformanceloss: time
medium

When A project has many secrets that share the same value.

Got The Insights page's duplicate check fetches every one of them at once, so it gets slower and heavier as the project grows.

Want The check shows a bounded page of duplicates, like the stale-secrets list.

steps
  1. Create a secrets project (blind indexing is on for new projects).
  2. Add 20 secrets that share one value; read the project's duplicated-secrets check: one group of 20.
  3. In a second project add 2000 secrets that share one value.
  4. Read that project's duplicated-secrets check (the Duplicated Secrets card on the project's Insights page reads it).
s8-project-duplicate-check-returns-every-secret: a project's duplicated-secrets insight is bounded
e2e/specs/secret-insights/s8-project-duplicate-check-returns-every-secret.spec.ts:40
lines
infisical/backend/src/services/secret-v2-bridge/secret-v2-bridge-dal.ts:1654
infisical/backend/src/services/secret-v2-bridge/secret-v2-bridge-dal.ts:1664
infisical/backend/src/ee/services/insights/insights-service.ts:687
infisical/backend/src/ee/services/insights/insights-service.ts:711
before the PRs testedno
The gateway search is cleared after opening a gateway and going back
s22-gateway-search-lost-on-back-1
gateways-relaysuiloss: time
low

When You search the gateway list, open a gateway, then go back.

Got The search box is empty and every gateway is listed again.

Want The list comes back as you left it, still searched.

steps
  1. Sign in as an organisation admin and open Networking on the Gateways tab
  2. Create a gateway, then type its name into "Search gateway..." so the list shows only that gateway
  3. Click the gateway's row to open its detail page
  4. Press the browser's Back button
s22-gateway-search-lost-on-back
e2e/specs/gateways-relays/s22-gateway-search-lost-on-back.spec.ts:5
lines
infisical/frontend/src/pages/organization/NetworkingPage/components/GatewayTab/GatewayTab.tsx:110
infisical/frontend/src/pages/organization/NetworkingPage/components/GatewayTab/GatewayTab.tsx:338
infisical/frontend/src/pages/organization/NetworkingPage/route.tsx:7
before the PRs testedno
Closing Create Proxy throws away the typed proxy name without asking
s4-create-proxy-draft-wiped-on-escape-1
uiloss: time
low

When

Got

Want

steps
  1. Sign in as an organization admin and open Agent Vault, Proxies
  2. Click Create Proxy and type a proxy name
  3. Close the form with Escape (Cancel, the X and a click outside do the same)
  4. Click Create Proxy again
lines
infisical/frontend/src/pages/agent-vault/AgentVaultProxiesPage/components/ProxyFormDialog.tsx:267
infisical/frontend/src/pages/agent-vault/AgentVaultProxiesPage/components/ProxyFormDialog.tsx:306
infisical/frontend/src/pages/agent-vault/AgentVaultProxiesPage/components/ProxyFormDialog.tsx:156
before the PRs testedno
A project's insights summary loads every due secret reminder at once
s9-project-summary-returns-every-due-reminder-1
secret-insightsperformanceloss: time
medium

When You open Insights for a project where many secrets have reminders due this week.

Got The summary loads every due reminder at once, so it gets slower and heavier as the project grows.

Want The summary shows a page of reminders with a total, like it does for stale secrets.

steps
  1. Create a secrets project and add 500 secrets.
  2. Give each secret a reminder repeating every day (due within 7 days).
  3. Read the project's insights summary with a stale-secrets page size of 50.
  4. Count the upcoming reminders in the answer.
s9-project-summary-returns-every-due-reminder: a project's insights summary bounds its due reminders
e2e/specs/secret-insights/s9-project-summary-returns-every-due-reminder.spec.ts:40
lines
infisical/backend/src/services/reminder/reminder-dal.ts:73
infisical/backend/src/ee/services/insights/insights-service.ts:199
infisical/backend/src/ee/services/insights/insights-service.ts:446
infisical/backend/src/ee/services/insights/insights-service.ts:467
before the PRs testedno
Honey token lists, details and counts scan every honey token of the instance
s9-honey-tokens-read-without-index-1
secret-insightsperformanceloss: time
medium

When You open a project's secrets overview, a honey token's details, or the project's insights.

Got Each of these looks through every honey token of every organisation on the instance, so they slow down as the instance grows.

Want Only the honey tokens of that folder or project are looked at.

steps
  1. Plant an AWS honey token in a secrets project
  2. Open the honey token's details (GET /api/v1/honey-tokens/:id?projectId=…) and the project's insights counts (GET /api/v1/insights/:projectId/secrets/counts)
  3. Count the statements the reads make and list the indexes on honey_tokens
s9-honey-tokens-read-without-index: honey tokens are found by folder through an index
e2e/specs/secret-insights/s9-honey-tokens-read-without-index.spec.ts:6
lines
infisical/backend/src/db/migrations/20260427174746_honey-tokens.ts:49
infisical/backend/src/ee/services/honey-token/honey-token-dal.ts:14
infisical/backend/src/ee/services/honey-token/honey-token-dal.ts:60
infisical/backend/src/ee/services/honey-token/honey-token-dal.ts:146
infisical/backend/src/ee/services/honey-token/honey-token-service.ts:1198
before the PRs testedno
A refused honey token leaves a live AWS IAM user and access key behind
s10-refused-honey-token-leaves-iam-user-1
secret-insightsbehaviourloss: time
medium

When You create an AWS honey token while your organisation's honey token stack is not deployed.

Got The product says the stack is not deployed, but it has already created an IAM user with an active access key in your AWS account, which it no longer knows about.

Want A refused honey token leaves nothing behind in your AWS account.

steps
  1. Configure the organisation's AWS honey token setup with an AWS connection and a deployed CloudFormation stack
  2. Delete that CloudFormation stack in AWS
  3. In a secrets project, create an AWS honey token
  4. The product refuses it: 'Honey token deployment verification failed. CloudFormation stack is not deployed.'
  5. List IAM users in the AWS account
a refused honey token create leaves no untracked IAM user with a live access key
e2e/specs/secret-insights/s10-refused-honey-token-leaves-iam-user.spec.ts:41
lines
infisical/backend/src/ee/services/honey-token/honey-token-service.ts:314
infisical/backend/src/ee/services/honey-token/honey-token-service.ts:342
infisical/backend/src/ee/services/honey-token/honey-token-service.ts:353
infisical/backend/src/ee/services/honey-token/honey-token-service.ts:359
infisical/backend/src/ee/services/honey-token/aws/honey-token-aws-fns.ts:108
infisical/backend/src/ee/services/honey-token/aws/honey-token-aws-fns.ts:110
before the PRs testedno
A finished EnvKey import is also reported as failed when its email can't be sent
s4-envkey-import-announced-failed-after-success-1
external-migrationbehaviourloss: decision
medium

When You import an EnvKey archive and the success email can't be sent.

Got You're told the import succeeded and then that it failed, yet every project from the archive was created.

Want One message that matches what was imported: the import succeeded.

steps
  1. In a fresh organisation, import an EnvKey archive from Integrations > External Migrations with the mail server up: one "Import Successful" notice and the archive's project appear.
  2. Start a second EnvKey import of another archive and wait for the "Infisical import started" email.
  3. Make the mail server stop answering after the second import has written its project (the test holds the product's create-project lock, pauses the mail server, then releases the lock).
  4. Wait for the import job to end, then open the notifications and the projects list.
An EnvKey import whose success mail cannot be sent is not also announced as failed
e2e/specs/external-migration/s4-envkey-import-announced-failed-after-success.spec.ts:100
lines
infisical/backend/src/services/external-migration/external-migration-queue.ts:174
infisical/backend/src/services/external-migration/external-migration-queue.ts:184
infisical/backend/src/services/external-migration/external-migration-queue.ts:192
infisical/backend/src/services/smtp/smtp-service.ts:226
infisical/backend/src/services/external-migration/external-migration-fns/import.ts:52
before the PRs testedno
Revoking a honey token can leave its decoy secrets planted in the folder
s11-honey-token-edit-during-revoke-plants-decoys-1
secret-insightsbehaviourloss: decision
low

When You revoke a honey token while someone else is saving new secret names for it.

Got Both succeed. The token shows as revoked, but two new decoy secrets with the new names stay in the folder.

Want The edit is refused once the token is revoked, and no decoy secrets are left.

steps
  1. Plant an AWS honey token in a secrets project.
  2. As one member, revoke the token.
  3. At the same time, as another member (or another tab), edit the token and change both secret names to new names, then save.
  4. Both actions report success.
  5. Open the folder the token was planted in.
a honey token revoked while its secret names are edited leaves no decoy secrets in its folder
e2e/specs/secret-insights/s11-honey-token-edit-during-revoke-plants-decoys.spec.ts:10
lines
infisical/backend/src/ee/services/honey-token/honey-token-service.ts:464
infisical/backend/src/ee/services/honey-token/honey-token-service.ts:547
infisical/backend/src/ee/services/honey-token/honey-token-service.ts:704
before the PRs testedno
A failed approver email makes a Vault import fire webhooks again for folders already imported
s5-vault-import-email-retry-refires-folder-webhook-1
external-migrationbehaviourloss: time
low

When You import Vault paths into folders where one needs approval, and the approvers' email fails once.

Got Folders that were imported straight away fire their webhooks and secret syncs a second time, with nothing changed.

Want Each imported folder fires its webhook and sync once per import.

steps
  1. Create a project with a webhook on folder /open and a change policy (one approver) on folder /guarded in the dev environment.
  2. Create a HashiCorp Vault connection to a Vault whose KV v2 mount kv holds kv/open and kv/guarded.
  3. Stop the mail server.
  4. Import kv/open and kv/guarded into the project at / with Keep Vault structure on.
  5. Wait until the API logs "Failed to notify approvers of Vault import change request" for /guarded, then start the mail server again.
  6. Wait for the approver email to go out on the retry and for the project's sync and webhook jobs to drain.
  7. Count the webhook deliveries for /open.
a Vault import whose approver email fails once fires the written folder's webhook once
e2e/specs/external-migration/s5-vault-import-email-retry-refires-folder-webhook.spec.ts:110
lines
infisical/backend/src/services/external-migration/external-migration-queue.ts:220
infisical/backend/src/services/external-migration/external-migration-queue.ts:249
infisical/backend/src/services/external-migration/external-migration-queue.ts:298
infisical/backend/src/services/secret/secret-queue.ts:616
before the PRs testedno
Setting up the authenticator app says "Invalid code" when the server fails, even for a correct code
s2-authenticator-enroll-server-failure-shown-as-invalid-code-1
behaviourloss: time
medium

When

Got

Want

steps
  1. Sign in as a user with no authenticator app configured
  2. Open Personal settings, Authentication tab, and click Set up on the Authenticator app row
  3. Have the authenticator enrol request fail on the server
  4. Type the correct six-digit code your authenticator app shows for the displayed key
lines
infisical/frontend/src/components/mfa/setup/VerifyStep.tsx:57
infisical/frontend/src/components/mfa/setup/VerifyStep.tsx:50
before the PRs testedno
Usage & Billing says "No products available" when the product list fails to load
s4-failed-product-catalog-shown-as-no-products-1
uiloss: time
medium

When

Got

Want

steps
  1. Sign in as an organization admin.
  2. Make the billing product catalog request fail with a server error (the subscription overview still loads).
  3. Open Usage & Billing for the organization.
lines
infisical/frontend/src/pages/organization/BillingV2Page/BillingV2Page.tsx:86
infisical/frontend/src/hooks/api/billingV2/queries.tsx:46
infisical/frontend/src/pages/organization/BillingV2Page/components/TabbedOverview.tsx:161
infisical/frontend/src/pages/organization/BillingV2Page/components/TabbedOverview.tsx:387
before the PRs testedno
Two sub-organizations can end up with the same slug when created at the same moment
s15-sub-org-slug-taken-twice-when-created-together-1
org-settingsbehaviourloss: time
low

When Two sub-organizations with the same slug are created at the same moment.

Got Both are created and the list shows two with one slug, though creating a taken slug one after the other is refused.

Want One is created and the other is told the slug already exists.

steps
  1. Sign in as an organization admin on the enterprise plan and open Organization Settings › Sub-Organizations.
  2. Create a sub-organization with slug s, then create another with slug s: the second is refused as 'Sub-organization with slug s already exists'.
  3. Create two sub-organizations with a fresh slug s2 at the same moment (the test holds the parent organization's row so both creates read before either writes).
  4. Search the Sub-Organizations list for s2.
two creates of one new sub-organization slug, read together, leave one sub-organization
e2e/specs/org-settings/s15-sub-org-slug-taken-twice-when-created-together.spec.ts:9
lines
infisical/backend/src/ee/services/sub-org/sub-org-service.ts:74
infisical/backend/src/ee/services/sub-org/sub-org-service.ts:83
infisical/backend/src/db/migrations/20251018061215_sub-org.ts:19
before the PRs testedno
Usage & Billing says the billing service is unreachable when the member is not allowed to view billing
s5-refused-billing-overview-reported-as-outage-1
uiloss: time
low

When

Got

Want

steps
  1. Sign in as an organization admin and open the organization.
  2. Make the subscription overview request (GET /api/v1/organizations/:id/billing/v2/overview) answer 403 Forbidden with the product's own error body.
  3. Open Usage & Billing.
lines
infisical/frontend/src/pages/organization/BillingV2Page/BillingV2Page.tsx:152
infisical/frontend/src/pages/organization/BillingV2Page/components/Overview.tsx:91
infisical/frontend/src/pages/organization/BillingV2Page/components/states/ErrorPanel.tsx:14
before the PRs testedno
The project template list gets slower with every template the organization adds
s12-template-list-reads-members-per-template-1
org-settingsperformanceloss: time
medium

When Your organization keeps adding project templates.

Got Every time the Project Templates tab loads, the app does extra work for each template, so the tab gets slower as the list grows.

Want The list loads with the same amount of work however many templates there are.

steps
  1. Sign in as an organization admin, create a new organization and 3 project templates in it
  2. List the organization's project templates (the Project Templates tab of Organization Settings) and count the database statements one list request makes
  3. Create 27 more project templates
  4. List them again and count the statements one list request makes
lines
infisical/backend/src/ee/services/project-template/project-template-service.ts:142
infisical/backend/src/ee/services/project-template/project-template-service.ts:136
infisical/backend/src/ee/services/project-template/project-template-user-membership-dal.ts:13
infisical/backend/src/ee/services/project-template/project-template-group-membership-dal.ts:13
infisical/backend/src/ee/services/project-template/project-template-identity-membership-dal.ts:15
infisical/backend/src/ee/routes/v1/project-template-router.ts:230
before the PRs testedno
The Sub-Organizations list cannot move past its first page
s7-sub-org-next-page-jumps-back-1
behaviourloss: time
medium

When

Got

Want

steps
  1. As an organization admin on the enterprise plan, create 21 sub-organizations whose names share a prefix.
  2. Open Organization Settings > Sub Organizations and search for the prefix: the table lists the first 20 and the footer reads "1 - 20 of 21".
  3. Click "Go to next page".
lines
infisical/frontend/src/pages/organization/SettingsPage/components/OrgSubOrgsTab/OrgSubOrgsTab.tsx:138
infisical/frontend/src/pages/organization/SettingsPage/components/OrgSubOrgsTab/OrgSubOrgsTab.tsx:153
infisical/frontend/src/pages/organization/SettingsPage/components/OrgSubOrgsTab/OrgSubOrgsTab.tsx:172
infisical/frontend/src/hooks/useResetPageHelper.ts:14
before the PRs testedno
Two project templates can end up with the same name in one organization
s11-template-name-taken-twice-when-saved-together-1
org-settingsbehaviourloss: nothing
low

When Two people in your organization save a project template under the same new name at the same moment.

Got Both are saved, and the list shows two templates with that name.

Want One is saved and the other is told the name already exists.

steps
  1. Sign in as an organization admin and open Organization Settings > Project Templates.
  2. Save a template named N, then save another named N: the second is refused as already existing.
  3. With a new name N2, send two template saves at the same moment so both check the name before either is written.
  4. Open the Project Templates list and search N2.
lines
infisical/backend/src/ee/services/project-template/project-template-service.ts:291
infisical/backend/src/ee/services/project-template/project-template-service.ts:427
infisical/backend/src/ee/services/project-template/project-template-service.ts:721
infisical/backend/src/db/migrations/20241101174939_project-templates.ts:10
before the PRs testedno
Secret sync destination policy switch has no name for screen readers
s17-secret-sync-policy-toggle-unnamed-1
org-settingsuiloss: nothing
low

When You use a screen reader on the Secrets Management product settings.

Got The Unique Secret Sync Destination Policy switch is announced only as a switch, with no name, and the Policies description text is slightly too faint.

Want The switch is announced with its policy's name, and the description is easy to read.

steps
  1. Sign in as an organisation admin.
  2. Open the organization's Secrets Management product settings and wait for the Policies card and its switches.
  3. Run an accessibility check (axe) on the page content, outside the sidebar and header.
secret management product settings toggles are named for screen readers
e2e/specs/org-settings/s17-secret-sync-policy-toggle-unnamed.spec.ts:16
lines
infisical/frontend/src/pages/organization/ProductSettingsPage/SecretsManagement/ProductSettingsPage.tsx:117
infisical/frontend/src/components/v3/generic/Card/Card.tsx:49
before the PRs testedno
Leaving Personal settings through "Back to organization" throws away an unsaved name without asking
s6-unsaved-name-dropped-on-back-to-organization-1
behaviourloss: time
low

When

Got

Want

steps
  1. Open Personal settings on the General tab.
  2. Type a new value in Name and do not click Save changes.
  3. Click the "Back to organization" link and wait for the organisation page to show.
  4. Press the browser's Back button.
lines
infisical/frontend/src/pages/user/PersonalSettingsPage/components/UserNameSection/UserNameSection.tsx:32
infisical/frontend/src/pages/user/PersonalSettingsPage/components/UserNameSection/UserNameSection.tsx:40
infisical/frontend/src/hooks/useNavigationBlocker.tsx:41
before the PRs testedno
The organization's Insights page has a view switch that points at missing content and a low-contrast "active leases" label
s12-org-insights-a11y-violations-1
secret-insightsuiloss: nothing
low

When A screen-reader or low-vision user opens the organization's Insights page.

Got The Needs Attention tab announces a panel that does not exist, and the "active leases" label is hard to read.

Want The view switch should refer to real content and the label should be readable.

steps
  1. Sign in as a member of an organization with a secrets project
  2. Open Secrets Management, then Insights at the organization level
  3. Run an axe accessibility scan on the page, leaving out the app shell
the organization's Insights page has no serious accessibility violations
e2e/specs/secret-insights/s12-org-insights-a11y-violations.spec.ts:10
lines
infisical/frontend/src/pages/organization/SecretInsightsPage/components/InsightsCard.tsx:195
infisical/frontend/src/pages/organization/SecretInsightsPage/components/SummaryCard.tsx:36
before the PRs testedno
Removing two server admins at once can leave the instance with none
s12-concurrent-server-admin-removals-leave-none-1
server-adminbehaviourloss: work
medium

When Two of the instance's server admins are removed at the same moment and they are the only two.

Got Both removals go through and no one is a server admin, locking everyone out of the Server Admin Console.

Want One removal goes through and the other is refused because it is the last server admin.

steps
  1. Sign in as the instance super admin; grant a second user server admin access so the instance has exactly two
  2. Remove one, then the other: the second is refused with 'Cannot remove the only server admin on this instance'
  3. Grant the second user server admin access again so there are two
  4. Remove both at the same moment (two admins or two tabs), with both requests reading the count of two before either removal is written
  5. Query the users table for rows with superAdmin = true
lines
infisical/backend/src/services/super-admin/super-admin-service.ts:867
infisical/backend/src/services/super-admin/super-admin-service.ts:871
infisical/backend/src/services/super-admin/super-admin-service.ts:878
before the PRs testedno
The org Insights project search runs off the screen on a phone
s13-org-insights-search-off-screen-on-phone-1
secret-insightsuiloss: time
low

When You open your organization's Insights page on a phone.

Got The project search box is pushed off the right edge of the screen; at the narrowest width only a sliver shows, and you cannot scroll sideways to it.

Want The search box fits on screen, under the view switch if needed.

steps
  1. Sign in as an admin of an organization with a secrets project
  2. Open the organization's Secrets Management Insights page on a screen 390 or 320 pixels wide
  3. Scroll down to the Insights card with the Needs Attention / All Projects switch
  4. Look for the Search projects field to the right of the switch
the Insights project search fits the screen at ${size.name} width (${size.width}px)
e2e/specs/secret-insights/s13-org-insights-search-off-screen-on-phone.spec.ts:8
lines
infisical/frontend/src/pages/organization/SecretInsightsPage/components/InsightsCard.tsx:194
infisical/frontend/src/pages/organization/SecretInsightsPage/components/InsightsCard.tsx:201
infisical/frontend/src/components/v3/generic/Card/Card.tsx:58
before the PRs testedno
Leaving Organization Settings throws away an unsaved organization name without asking
s8-unsaved-org-name-dropped-on-leave-1
behaviourloss: time
low

When

Got

Want

steps
  1. Sign in as an organization admin and open Organization Settings, General tab.
  2. Type a new value in Organization Name and do not click Save.
  3. Click the SSO link in the organization sidebar and wait for the SSO page.
  4. Press the browser's Back button.
lines
infisical/frontend/src/pages/organization/SettingsPage/components/OrgNameChangeSection/OrgNameChangeSection.tsx:57
infisical/frontend/src/pages/organization/SettingsPage/components/OrgNameChangeSection/OrgNameChangeSection.tsx:67
infisical/frontend/src/hooks/useNavigationBlocker.tsx:41
before the PRs testedno
Cancelling "Add Sub Org" throws away the typed name without asking
s9-add-sub-org-cancel-discards-name-1
behaviourloss: time
low

When

Got

Want

steps
  1. Sign in as an organization admin on the enterprise plan and open Organization Settings > Sub-Organizations.
  2. Click Add Sub Org.
  3. Type a Display Name.
  4. Click Cancel.
  5. Click Add Sub Org again.
lines
infisical/frontend/src/components/organization/NewSubOrganizationModal/NewSubOrganizationModal.tsx:61
infisical/frontend/src/components/organization/NewSubOrganizationModal/NewSubOrganizationModal.tsx:138
before the PRs testedno
Cancelling Add Template throws away the typed template name without asking
s8-add-template-cancel-discards-name-1
uiloss: time
low

When

Got

Want

steps
  1. Sign in as an organization admin and open the organization's KMS product settings, where the project templates are listed
  2. Click Add Template and type a template name
  3. Press Cancel
  4. Click Add Template again
lines
infisical/frontend/src/pages/organization/ProductSettingsPage/project-templates/ProjectTemplateDetailsModal.tsx:110
infisical/frontend/src/pages/organization/ProductSettingsPage/project-templates/ProjectTemplateDetailsModal.tsx:135
before the PRs testedno
An unsaved session length is lost without warning when leaving the Security tab
s18-unsaved-session-length-dropped-on-leave-1
org-settingsbehaviourloss: time
low

When You type a new session length on the Security tab and click another page in the sidebar before saving.

Got Nothing warns you. When you come back, the field shows the old length and what you typed is gone.

Want The app warns you before leaving, or keeps what you typed.

steps
  1. Sign in as an organization admin and open Organization Settings, Security tab.
  2. Type 17 in the user session Expiration Value field; do not save.
  3. Click the SSO link in the organization sidebar.
  4. Press the browser Back button.
lines
infisical/frontend/src/pages/organization/SettingsPage/components/OrgSecurityTab/OrgUserAccessTokenLimitSection.tsx:66
before the PRs testedno
A saved session length cannot be changed back to its old value without a reload
s10-session-length-revert-after-save-blocked-1
behaviourloss: decision
medium

When

Got

Want

steps
  1. Sign in as an organization admin and open Organization Settings, Security tab.
  2. Note the Session Length Expiration Value (for example 90 Days).
  3. Change Expiration Value to 45 and click Save; the success message appears.
  4. Type the previous value (90) back into Expiration Value.
lines
infisical/frontend/src/pages/organization/SettingsPage/components/OrgSecurityTab/OrgUserAccessTokenLimitSection.tsx:76
infisical/frontend/src/pages/organization/SettingsPage/components/OrgSecurityTab/OrgUserAccessTokenLimitSection.tsx:171
before the PRs testedno
Organization pages crash on the way to sign-in once the session has expired
s19-expired-session-click-crashes-org-shell-1
org-settingsuiloss: nothing
low

When You come back to an organization page after your session has ended and click a link.

Got You land on the sign-in page, but the page crashes on the way there.

Want You are taken to sign in cleanly.

steps
  1. Sign in as an organization admin and open Organization Settings, Security tab.
  2. Let the session end: sign out in another tab of the same browser, and wait past the access token's lifetime with the tab left open.
  3. In the open tab, click the organization's SSO link in the sidebar.
an expired session clicking to another organization page goes to sign-in without a crash
e2e/specs/org-settings/s19-expired-session-click-crashes-org-shell.spec.ts:5
lines
infisical/frontend/src/components/v3/platform/RootCommandMenu/RootCommandMenu.tsx:1157
infisical/frontend/src/components/v3/platform/RootCommandMenu/RootCommandMenu.tsx:1116
infisical/frontend/src/pages/root.tsx:47
infisical/frontend/src/pages/middlewares/authenticate.tsx:43
before the PRs testedno
Cancelling the HashiCorp Vault import throws away the typed Vault URL without asking
s6-vault-url-lost-on-cancel-1
external-migrationbehaviourloss: time
low

When You start a HashiCorp Vault import, type the Vault URL, and press Cancel.

Got The dialog closes straight away and the URL is gone when you come back, with no warning.

Want You are asked before your typed details are discarded.

steps
  1. Sign in as an organization admin and open Integrations > External Migrations.
  2. Click Import, then pick HCP Vault.
  3. Type a Vault URL.
  4. Click Cancel.
  5. Click Import again and pick HCP Vault.
s6-vault-url-lost-on-cancel
e2e/specs/external-migration/s6-vault-url-lost-on-cancel.spec.ts:7
lines
infisical/frontend/src/pages/organization/SettingsPage/components/ExternalMigrationsTab/components/VaultPlatformModal.tsx:373
infisical/frontend/src/pages/organization/SettingsPage/components/ExternalMigrationsTab/components/SelectImportFromPlatformModal.tsx:45
before the PRs testedno
Pressing Esc in the Add Honey Token wizard throws away the typed details without asking
s5-honey-token-wizard-discarded-on-esc-1
uiloss: time
low

When

Got

Want

steps
  1. Sign in and open a secret-management project's overview
  2. Choose Add New, then Add Honey Token
  3. Pick the Development environment and press Next
  4. Fill the secret name for the Access Key ID and press Next
  5. Type a name on the Details step
  6. Press Esc
  7. Open Add Honey Token again and walk back to the Details step
lines
infisical/frontend/src/components/honey-tokens/CreateHoneyTokenModal.tsx:67
infisical/frontend/src/components/honey-tokens/CreateHoneyTokenModal.tsx:53
infisical/frontend/src/components/honey-tokens/forms/HoneyTokenForm.tsx:88
before the PRs testedno
Leaving the admin General page throws away unsaved allowed email domains without asking
s14-unsaved-signup-domains-dropped-on-leave-1
server-adminbehaviourloss: time
low

When You type new allowed email domains on the admin General page and open another admin page before saving.

Got Nothing warns you. When you come back, the domains you typed are gone.

Want The console warns you before leaving, or keeps what you typed.

steps
  1. Sign in as the instance super admin and open the Server Console's General page.
  2. Type a new domain in Allowed email domains and do not click Save.
  3. Click Encryption in the admin navigation and wait for the Encryption page.
  4. Press the browser's Back button.
s14-unsaved-signup-domains-dropped-on-leave
e2e/specs/server-admin/s14-unsaved-signup-domains-dropped-on-leave.spec.ts:8
lines
infisical/frontend/src/pages/admin/GeneralPage/components/GeneralPageForm.tsx:112
infisical/frontend/src/pages/admin/GeneralPage/components/GeneralPageForm.tsx:121
infisical/frontend/src/hooks/useNavigationBlocker.tsx:13
before the PRs testedno
Leaving the admin Environment Variables page throws away an unsaved override without asking
s14-unsaved-env-override-dropped-on-leave-1
server-adminbehaviourloss: time
low

When You type an environment override, don't save, open Caching and come back.

Got Nothing warns you. The override field is empty again and your typing is gone.

Want The page warns you before leaving, or keeps what you typed.

steps
  1. Sign in as the super admin and open the admin console's Environment Variables page.
  2. Open the Heroku group and type a value into INF_APP_CONNECTION_HEROKU_OAUTH_CLIENT_ID; do not click Save.
  3. Click Caching in the admin console's navigation and wait for the Caching page.
  4. Press the browser's Back button and open the Heroku group again.
an unsaved Heroku override survives a trip to Caching and back, or a warning asks first
e2e/specs/server-admin/s14-unsaved-env-override-dropped-on-leave.spec.ts:8
lines
infisical/frontend/src/pages/admin/EnvironmentPage/components/EnvironmentPageForm.tsx:129
infisical/frontend/src/pages/admin/EnvironmentPage/components/EnvironmentPageForm.tsx:161
infisical/frontend/src/hooks/useNavigationBlocker.tsx:41
before the PRs testedno
An environment variable search that matches nothing leaves a blank card
s15-env-variable-search-no-match-blank-1
server-adminuiloss: nothing
low

When You search the Environment Variables page for something no variable matches.

Got The list of variables just disappears and the card is blank, with no message.

Want The card says nothing matched your search.

steps
  1. Sign in as the instance super admin on an instance whose setup is finished
  2. Open the Server Console and go to Environment Variables
  3. See the Overrides card list its variable groups (Heroku among them)
  4. Type zzheal-no-such-variable-qq into Search environment variables
s15-env-variable-search-no-match-blank a no-match search says nothing matched
e2e/specs/server-admin/s15-env-variable-search-no-match-blank.spec.ts:6
lines
infisical/frontend/src/pages/admin/EnvironmentPage/components/EnvironmentPageForm.tsx:217
infisical/frontend/src/pages/admin/EnvironmentPage/components/EnvironmentPageForm.tsx:136
before the PRs testedno
Searching the admin Users list from page 2 hides the matching user
s7-admin-user-search-from-page-two-shows-empty-page-1
behaviourloss: time
medium

When

Got

Want

steps
  1. Sign in as the instance super admin on an instance with more than 20 users and open the Server Console's Resources overview, Users tab
  2. Go to the next page of users (the pager reads 21 - ...)
  3. Type the seeded admin's email (seed-admin-a@heal-sbx.test) into Search users
lines
infisical/frontend/src/pages/admin/ResourceOverviewPage/components/UserIdentitiesTable.tsx:347
infisical/frontend/src/pages/admin/ResourceOverviewPage/components/UserIdentitiesTable.tsx:349
infisical/frontend/src/pages/admin/ResourceOverviewPage/components/UserIdentitiesTable.tsx:358
infisical/frontend/src/pages/admin/ResourceOverviewPage/components/EmailDomainsTable.tsx:218
before the PRs testedno
Signed-out visitors wait an extra second before being sent to sign-in
s14-refused-session-refresh-retried-1
sign-inperformanceloss: time
low

When You open a protected page while signed out.

Got The app checks for your session again after it already knows you have none, and the sign-in page appears about a second late.

Want You are sent to sign-in as soon as the app knows you have no session.

steps
  1. Open a fresh browser with no Infisical session
  2. Go directly to /personal-settings
  3. Watch the requests to the session refresh endpoint until the login page shows
a signed-out visit is sent to sign-in without retrying the refused refresh
e2e/specs/sign-in/s14-refused-session-refresh-retried.spec.ts:7
lines
infisical/frontend/src/pages/middlewares/authenticate.tsx:18
infisical/frontend/src/hooks/api/auth/refresh.ts:23
infisical/frontend/src/hooks/api/reactQuery.tsx:350
before the PRs testedno
A KMS project's connection links to a "page not found" screen
s9-kms-project-connection-link-not-found-1
app-connectionsbehaviourloss: time
low

When A connection was added to a KMS project through the API, and you open it from the organisation's connections list.

Got Clicking the project name opens a 'page not found' screen.

Want The link opens a page that shows the connection, or the connection could not be added there.

steps
  1. As an org admin, create a KMS project.
  2. Through the public API, create an AWS app connection with access-key credentials and projectId set to that KMS project; it is accepted.
  3. Open the organisation's Integrations page, App Connections tab, and switch the view to All.
  4. Find the connection and click its project name in the project column.
s9-kms-project-connection-link-not-found
e2e/specs/app-connections/s9-kms-project-connection-link-not-found.spec.ts:9
lines
infisical/frontend/src/pages/organization/AppConnections/AppConnectionsPage/components/AppConnectionRow.tsx:205
infisical/frontend/src/pages/organization/AppConnections/AppConnectionsPage/components/AppConnectionRow.tsx:209
infisical/frontend/src/helpers/project.ts:197
infisical/frontend/src/routes.ts:30
infisical/backend/src/services/app-connection/app-connection-service.ts:580
infisical/backend/src/services/app-connection/app-connection-fns.ts:394
before the PRs testedno
A freshly emailed password-setup link can be refused as expired
s5-concurrent-setup-emails-leave-two-live-tokens-1
account-recoverybehaviourloss: time
low

When You ask the product to email you a password-setup link twice in quick succession.

Got You get two emails, but one of the links says the token is expired or invalid.

Want Every setup link the product just emailed you works, or you simply get one link.

steps
  1. Sign in as an account and request a password-setup email once, so one live setup token exists.
  2. Request two password-setup emails for the same account at the same instant (forced in the test by holding the existing token row so both requests read before either writes).
  3. Count the account's live setup tokens: two remain, not one.
  4. Open the link from the earlier of the two emails: it is refused as expired or invalid, because only one of the two tokens is the one the server will validate.
  5. The same happens on the MFA sign-in step: with an organization that enforces email MFA, reach the MFA code prompt and send a new code twice at the same instant; two emailMfa codes stay live for the user instead of one.
  6. The same happens on an organization invitation: invite a fresh email, then resend that member's invitation twice at the same instant from Access Management; two organizationInvitation codes stay live for the invitee instead of one.
lines
infisical/backend/src/services/auth-token/auth-token-service.ts:121
infisical/backend/src/services/auth-token/auth-token-service.ts:122
infisical/backend/src/services/auth-token/auth-token-service.ts:123
infisical/backend/src/services/auth-token/auth-token-service.ts:206
infisical/backend/src/services/auth/auth-password-service.ts:110
infisical/backend/src/services/auth/auth-login-service.ts:162
infisical/backend/src/services/auth/auth-login-service.ts:646
infisical/backend/src/services/org/org-service.ts:1183
infisical/backend/src/db/migrations/20231129072939_auth-token.ts:8
before the PRs testedno
Inviting someone to a project fails when their email has capital letters
s14-capitalised-invitee-email-refused-1
project-access-controlbehaviourloss: time
low

When You add a new person to a project by typing their email with a capital letter.

Got The form offers to invite them, then refuses with 'username must be lowercase' and nobody is invited.

Want The person is invited under their email, whatever its case.

steps
  1. Sign in as a project admin and open the project's Access Control page, Users tab
  2. Click Add Users to Project
  3. Type a new email with capital letters, e.g. New.Person@Example.com, and pick the offered 'Invite "New.Person@Example.com"' option
  4. Pick the Member role
  5. Click Add Members
a new invitee typed with capital letters is added to the project
e2e/specs/project-access-control/s14-capitalised-invitee-email-refused.spec.ts:12
lines
infisical/frontend/src/pages/project/AccessControlPage/components/MembersTab/components/AddMemberModal.tsx:328
infisical/frontend/src/pages/project/AccessControlPage/components/MembersTab/components/AddMemberModal.tsx:334
infisical/frontend/src/pages/project/AccessControlPage/components/MembersTab/components/AddMemberModal.tsx:202
infisical/backend/src/server/routes/v1/project-membership-router.ts:379
before the PRs testedno
A failed organization open after a correct MFA code ends in a false lockout
s15-org-open-failure-after-mfa-locks-out-1
sign-inbehaviourloss: time
medium

When Your MFA code is accepted but your organization then fails to open.

Got You stay on the code screen. Each retry says your code is invalid and counts down, and after five you are signed out as if locked.

Want Retrying opens your organization, since your code was already accepted.

steps
  1. Sign up a new account and turn on enforced MFA for its organization.
  2. Sign in once with the emailed code and finish the recovery-codes step, so MFA is set up.
  3. Sign out, sign in again with email and password, and reach the MFA code prompt.
  4. Make opening the organization fail once with a server error, then enter the correct emailed code.
  5. After the error notice, with the server healthy again, press Try again on the code prompt.
s15-org-open-failure-after-mfa-locks-out
e2e/specs/sign-in/s15-org-open-failure-after-mfa-locks-out.spec.ts:23
lines
infisical/frontend/src/pages/auth/SelectOrgPage/SelectOrgPage.tsx:261
infisical/frontend/src/pages/auth/SelectOrgPage/SelectOrgPage.tsx:466
infisical/frontend/src/components/auth/Mfa.tsx:117
infisical/frontend/src/components/auth/Mfa.tsx:136
infisical/backend/src/server/routes/v2/mfa-router.ts:72
#8620
reviewed by greptile, cubic, coderabbit
no
Adding users to a project that needs an upgrade reports success though nobody is added
s14-legacy-project-member-add-reports-success-1
project-access-controlbehaviourloss: decision
medium

When You add a user to an older project that still needs an upgrade.

Got You see an upgrade error and a success message at once, and the dialog closes, but nobody was added.

Want Only the upgrade error shows and the dialog stays open.

steps
  1. Have a secret-management project that is still at version 1 (a legacy project not yet upgraded).
  2. Open the project's Access Control page, Users tab.
  3. Click Add Users to Project, pick an organisation member not in the project and a role.
  4. Click Add Members.
only the upgrade error is shown, no success message
e2e/specs/project-access-control/s14-legacy-project-member-add-reports-success.spec.ts:36
lines
infisical/frontend/src/pages/project/AccessControlPage/components/MembersTab/components/AddMemberModal.tsx:170
infisical/frontend/src/pages/project/AccessControlPage/components/MembersTab/components/AddMemberModal.tsx:210
infisical/frontend/src/pages/project/AccessControlPage/components/MembersTab/components/AddMemberModal.tsx:214
infisical/frontend/src/pages/secret-manager/OverviewPage/components/InviteMembersModal/InviteMembersModal.tsx:121
before the PRs testedno
A refused SAML sign-in ends on a raw error page instead of the app
s27-refused-saml-assertion-shows-raw-json-1
sso-loginbehaviourloss: time
low

When Your SAML identity provider turns down your sign-in (an unsigned or expired assertion, a key mismatch).

Got You are left on a bare page of code-like text saying Saml authentication failed, outside the app, with no way back.

Want You land on Infisical's sign-in error page with a clear message and a way back to sign in.

steps
  1. As an organization admin, configure and activate SAML single sign-on for the organization (the SAML section of the organization SSO page, or POST /api/v1/sso/config with an entryPoint, issuer and self-signed cert).
  2. Sign out, and have the SAML provider post back an assertion that fails validation (unsigned, signed by another key, or expired) to the sign-in return address /api/v1/sso/saml2/<configId> as an application/x-www-form-urlencoded form.
  3. Observe the resulting page.
s27-refused-saml-assertion-shows-raw-json
e2e/specs/sso-login/s27-refused-saml-assertion-shows-raw-json.spec.ts:36
lines
infisical/backend/src/ee/routes/v1/saml-router.ts:365
infisical/backend/src/ee/routes/v1/saml-router.ts:371
infisical/backend/src/ee/routes/v1/saml-router.ts:372
before the PRs testedno
A machine identity's alerts look absent when they fail to load
s16-identity-alerts-failure-shown-as-no-alerts-1
machine-identitiesuiloss: decision
medium

When You open a machine identity and its alerts cannot be loaded.

Got The Alerts button looks like the identity has no alerts, with no sign anything went wrong.

Want The page tells you the alerts could not be loaded.

steps
  1. Sign in as an organization admin and open a machine identity's detail page.
  2. Create an authentication expiry alert for it; the header shows the Alerts button with the count 1.
  3. Make the alerts load fail with a server error (the query retries once, so both attempts fail).
  4. Reload the identity's detail page and wait for it to settle.
s16-identity-alerts-failure-shown-as-no-alerts
e2e/specs/machine-identities/s16-identity-alerts-failure-shown-as-no-alerts.spec.ts:13
lines
infisical/frontend/src/views/Alerts/AlertAction.tsx:74
infisical/frontend/src/views/Alerts/AlertAction.tsx:92
infisical/frontend/src/views/Alerts/AlertAction.tsx:262
infisical/frontend/src/pages/organization/IdentityDetailsByIDPage/components/IdentityAlertAction.tsx:11
infisical/frontend/src/pages/project/IdentityDetailsByIDPage/components/ProjectIdentityAlertAction.tsx:16
#8526
reviewed by greptile, cubic, coderabbit
no
An incomplete AWS Parameter Store create link opens a blank screen
s12-create-page-without-authorization-blank-1
legacy-integrationsuiloss: time
low

When You open an AWS Parameter Store create link that is missing part of its address.

Got The screen stays completely blank, with no message and no way back.

Want An error page that says the link is incomplete and lets you go back.

steps
  1. Sign in and open a Secrets Management project
  2. Open the project's AWS Parameter Store create page by its address, leaving off the integrationAuthId part of the link
  3. Wait for the page to finish loading
the AWS Parameter Store create page without an authorization id shows a page, not a blank screen
e2e/specs/legacy-integrations/s12-create-page-without-authorization-blank.spec.ts:6
lines
infisical/frontend/src/pages/organization/layout.tsx:15
infisical/frontend/src/pages/secret-manager/integrations/AwsParameterStoreConfigurePage/route.tsx:17
before the PRs testedno
An earlier project's data source links to an empty findings list
s16-earlier-project-data-source-findings-link-empty-1
secret-scanningbehaviourloss: decision
medium

When You open a data source that belongs to your organisation's earlier Secret Scanning project.

Got Its repository shows '1 Secret Detected', but the link opens an empty findings list, as if the leak were gone.

Want The link shows the leak, or the page says the data source is in another project.

steps
  1. In an organisation with two Secret Scanning projects, be an admin of both; the earlier one has a GitHub data source whose repository has one unresolved leak
  2. With no earlier project chosen in the browser (a fresh browser, or after 'Go to Secret Scanning'), open that data source's page by its link
  3. On the repository row, follow '1 Secret Detected'
s16 a data source of an earlier Secret Scanning project links to that project's findings
e2e/specs/secret-scanning/s16-earlier-project-data-source-findings-link-empty.spec.ts:77
lines
infisical/frontend/src/pages/secret-scanning/layout.tsx:35
infisical/frontend/src/pages/secret-scanning/SecretScanningDataSourceByIdPage/SecretScanningDataSourceByIdPage.tsx:34
infisical/frontend/src/pages/secret-scanning/SecretScanningDataSourceByIdPage/components/SecretScanningResourceRow.tsx:149
infisical/frontend/src/helpers/secretScanningActiveProject.ts:25
#8568
reviewed by greptile, cubic, coderabbit
no
Fixing a failing log stream's endpoint is undone: it shows Failing again
s10-fixed-stream-flagged-by-old-delivery-1
audit-logsbehaviourloss: decision
medium

When You fix a failing log stream's endpoint while its last retry is still on its way to the old one.

Got The stream shows healthy, then a moment later Failing again with the old endpoint's error.

Want The stream stays healthy, since the error came from the endpoint you replaced.

steps
  1. Create a Custom audit log stream whose endpoint is down, and let its deliveries fail until the stream shows Failing.
  2. Let another batch of events reach its last delivery attempt, while the old endpoint is slow to answer (under 30 s).
  3. While that delivery is in flight, edit the stream, set a working endpoint URL and save: the stream shows healthy.
  4. The old endpoint answers the in-flight batch with an error.
  5. Read the stream list again.
A delivery in flight to a log stream's replaced endpoint does not mark the fixed stream failing
e2e/specs/audit-logs/s10-fixed-stream-flagged-by-old-delivery.spec.ts:34
lines
infisical/backend/src/ee/services/audit-log-stream-outbox/audit-log-stream-outbox-service.ts:256
infisical/backend/src/ee/services/audit-log-stream-outbox/audit-log-stream-outbox-service.ts:416
infisical/backend/src/ee/services/audit-log-stream-outbox/audit-log-stream-outbox-dal.ts:309
infisical/backend/src/ee/services/audit-log-stream/audit-log-stream-service.ts:202
#8526
reviewed by greptile, cubic, coderabbit
no
Screen readers hear an unnamed button beside a temporary project role
s15-temporary-role-icon-has-no-name-1
project-access-controluiloss: nothing
low

When A project member has a temporary role and you browse the Users list with a screen reader.

Got Beside the role you reach a button that is announced with no name, so you can't tell it marks temporary or expired access.

Want The icon is announced as "Temporary access" or "Access expired".

steps
  1. Give a project member a permanent Viewer role plus a Member role marked temporary for 2 hours
  2. Open the project's Access Management page, Users tab
  3. Search for that member
  4. Tab to or inspect the clock icon beside the role badge with a screen reader
a temporary project role's icon in the Users table is named for screen readers
e2e/specs/project-access-control/s15-temporary-role-icon-has-no-name.spec.ts:12
lines
infisical/frontend/src/pages/project/AccessControlPage/components/MembersTab/components/MembersTable.tsx:429
infisical/frontend/src/pages/project/AccessControlPage/components/MembersTab/components/MembersTable.tsx:484
before the PRs testedno
Starting sign-up with no connection says "Bad Request", not that the network is down
s3-offline-signup-shown-as-bad-request-1
behaviourloss: time
low

When

Got

Want

steps
  1. Open the sign-up page while signed out
  2. Lose the network connection to the server
  3. Type a new email address and choose Continue with Email
  4. Read the notification that appears
lines
infisical/frontend/src/hooks/api/reactQuery.tsx:323
infisical/frontend/src/hooks/api/reactQuery.tsx:327
infisical/frontend/src/pages/organization/SettingsPage/components/OrgSsoTab/OIDCModal.tsx:211
before the PRs testedno
Signing in with SAML or OIDC by an email with no SSO shows a raw error page
s26-unknown-saml-domain-shows-raw-json-1
sso-loginuiloss: time
low

When You try to sign in with SAML using a work email whose company has no SAML set up.

Got You leave Infisical and land on a page of raw code-like text.

Want Stay on the sign-in screen with a message that SAML is not set up for that email.

steps
  1. Sign out and open the SAML login page (/login/saml) or the OIDC login page (/login/oidc)
  2. Type a work email whose domain has no SSO, e.g. nobody@no-such-domain-heal.example — or open the SAML login page through a link naming an organization slug no organization has (/login/saml?organizationSlug=<unknown slug>)
  3. Press Continue with SAML (or Continue with OIDC)
lines
infisical/frontend/src/pages/auth/LoginSsoPage/LoginSsoPage.tsx:62
infisical/frontend/src/pages/auth/LoginSsoPage/LoginSsoPage.tsx:70
infisical/backend/src/ee/routes/v1/saml-router.ts:86
infisical/backend/src/ee/routes/v1/saml-router.ts:285
infisical/backend/src/ee/routes/v1/oidc-router.ts:108
before the PRs testedno
The organization list looks empty when it fails to load
s10-select-org-load-failure-shown-as-empty-1
sign-inuiloss: time
medium

When You open the choose-organization screen and your organizations cannot be loaded.

Got The screen says "No organizations found", as if you belonged to none, with no error and no retry.

Want The screen says your organizations could not be loaded and lets you try again.

steps
  1. Sign in as a user who belongs to an organization.
  2. Open the choose-organization screen (/login/select-organization?force=true) and see the organization listed.
  3. Make the organization list request fail with a server error.
  4. Reload the screen.
lines
infisical/frontend/src/pages/auth/SelectOrgPage/SelectOrgPage.tsx:139
infisical/frontend/src/pages/auth/SelectOrgPage/SelectOrgPage.tsx:196
infisical/frontend/src/pages/auth/SelectOrgPage/SelectOrgPage.tsx:352
infisical/frontend/src/hooks/api/organization/queries.tsx:85
before the PRs testedno
A failed email-domain load on the SSO page says the organization has no domains
s27-failed-email-domain-load-reads-as-none-1
sso-loginbehaviourloss: decision
medium

When You open the SSO settings and your email domains fail to load.

Got The page says you have no email domains and must verify one before adding a provider, with no sign of an error.

Want The page says the domains could not be loaded.

steps
  1. Sign in as an organization admin on the enterprise plan.
  2. Have the email domains request fail with a server error (both the first try and the automatic retry).
  3. Open Organization Settings > SSO & Provisioning.
lines
infisical/frontend/src/hooks/api/emailDomains/queries.tsx:12
infisical/frontend/src/pages/organization/SettingsPage/components/OrgSsoTab/OrgEmailDomainsSection/OrgEmailDomainsTable.tsx:108
infisical/frontend/src/pages/organization/SettingsPage/components/OrgSsoTab/OrgSsoTab.tsx:163
before the PRs testedno
The OIDC sign-in screen forgets the typed work email when you come back
s14-oidc-work-email-lost-after-back-1
behaviourloss: time
low

When

Got

Want

steps
  1. Signed out, open the OIDC login page (/login/oidc).
  2. Type a work email (jane.doe@acme-heal.example) into Work Email; do not continue.
  3. Click the Back to login arrow and wait for the main login page to show.
  4. Press the browser's back button to return to the OIDC login page.
lines
infisical/frontend/src/pages/auth/LoginSsoPage/LoginSsoPage.tsx:41
infisical/frontend/src/pages/auth/LoginSsoPage/LoginSsoPage.tsx:94
infisical/frontend/src/pages/auth/LoginSamlPage/LoginSamlPage.tsx:3
before the PRs testedno
AWS S3 buckets in newer regions cannot be connected as S3-Compatible Storage
s10-newer-aws-region-s3-endpoint-refused-1
app-connectionsbehaviourloss: time
medium

When You connect an AWS S3 bucket in a newer region, such as Calgary, as S3-Compatible Storage.

Got The form accepts the endpoint, then saving says it is not the S3 endpoint of a supported provider, though it is exactly the AWS S3 endpoint the message lists.

Want Any AWS S3 region's endpoint is accepted, and only wrong keys are refused.

steps
  1. Sign in as an organisation admin and open Integrations, App Connections tab
  2. Add a connection, pick S3-Compatible Storage, access key method
  3. Enter the endpoint https://s3.ca-west-1.amazonaws.com and any access key id and secret
  4. Save the connection
an S3-Compatible Storage connection to AWS S3 in ca-west-1 is not refused as an unsupported endpoint
e2e/specs/app-connections/s10-newer-aws-region-s3-endpoint-refused.spec.ts:8
lines
infisical/backend/src/services/app-connection/s3-compatible/s3-compatible-connection-schemas.ts:52
infisical/backend/src/services/app-connection/app-connection-enums.ts:96
infisical/backend/src/services/app-connection/s3-compatible/s3-compatible-connection-fns.ts:43
infisical/frontend/src/helpers/appConnections.ts:695
infisical/frontend/src/pages/organization/AppConnections/AppConnectionsPage/components/AppConnectionForm/S3CompatibleConnectionForm.tsx:58
#8485
reviewed by greptile, cubic, coderabbit
no
A Portainer connection that can't list environments looks like it has none
s12-portainer-environment-list-failure-shows-no-options-1
secret-syncsuiloss: time
medium

When You add a Portainer sync with a connection whose Portainer instance refuses its token.

Got The Environment list says "No options found." and nothing tells you anything went wrong.

Want The form tells you the environments could not be loaded and why.

steps
  1. Create a Portainer app connection in the project's organisation whose Portainer instance later refuses the API token (for example, the token is revoked in Portainer after the connection was saved).
  2. Open the project's Integrations page, Secret Syncs tab, and click Add Sync.
  3. Pick Portainer, continue past the Source step.
  4. On the Destination step, pick that Portainer connection.
  5. Open the Environment picker.
s12-portainer-environment-list-failure-shows-no-options
e2e/specs/secret-syncs/s12-portainer-environment-list-failure-shows-no-options.spec.ts:22
lines
infisical/frontend/src/components/secret-syncs/forms/SecretSyncDestinationFields/PortainerSyncFields.tsx:36
infisical/frontend/src/components/secret-syncs/forms/SecretSyncDestinationFields/PortainerSyncFields.tsx:41
infisical/frontend/src/components/secret-syncs/forms/SecretSyncDestinationFields/PortainerSyncFields.tsx:83
infisical/frontend/src/hooks/api/appConnections/portainer/queries.tsx:28
infisical/backend/src/services/app-connection/portainer/portainer-connection-fns.ts:214
#7670
reviewed by greptile, cubic, coderabbit
no
Saving Universal Auth settings undoes another admin's newer use limit and lockout
s13-stale-universal-auth-dialog-reverts-restriction-1
identity-authbehaviourloss: work
medium

When You have a machine identity's Universal Auth edit dialog open while another admin tightens its token use limit and lockout.

Got You change only the token TTL and save. Their use limit and lockout are back to the old values, and nobody is told.

Want Their restriction stays, or you are told the settings changed since you opened the dialog.

steps
  1. As an org admin, create a machine identity and add Universal Auth with Access Token TTL 3600, no token use limit and lockout threshold 3.
  2. Open the identity's page, open the Universal Auth method and click [Edit]; the dialog shows TTL 3600.
  3. As another admin (or through the API), save a token use limit of 5 and a lockout threshold of 10 on that identity's Universal Auth (accepted).
  4. In the still-open dialog, change only Access Token TTL to 7200 and click [Update] (accepted).
  5. Read the identity's Universal Auth settings.
s13-stale-universal-auth-dialog-reverts-restriction
e2e/specs/identity-auth/s13-stale-universal-auth-dialog-reverts-restriction.spec.ts:43
lines
infisical/frontend/src/pages/organization/AccessManagementPage/components/OrgIdentityTab/components/IdentitySection/IdentityUniversalAuthForm.tsx:148
infisical/frontend/src/pages/organization/AccessManagementPage/components/OrgIdentityTab/components/IdentitySection/IdentityUniversalAuthForm.tsx:203
infisical/backend/src/services/identity-ua/identity-ua-service.ts:660
infisical/backend/src/server/routes/v1/identity-universal-auth-router.ts:311
before the PRs testedno
Create CA wizard says no HSM connectors exist when they failed to load
s9-hsm-connector-list-failure-shown-as-none-1
certificate-authoritiesbehaviourloss: decision
medium

When You create a CA with an HSM key and the list of HSM connectors can't be loaded.

Got The picker says no HSM connectors are configured and tells you to add one, with no sign anything failed.

Want The picker says the connectors couldn't be loaded and lets you try again.

steps
  1. Sign in as an admin of a Certificate Management project on a plan that includes HSM.
  2. Open Certificate Authorities and click Create Internal CA.
  3. Fill the name, continue, fill the common name and organisation, continue to the Key & Validity step.
  4. Have the HSM connector list request fail with a server error (and its one retry).
  5. Pick HSM as the key source and open the HSM Connector picker.
a failed HSM connector list is not presented as 'No HSM Connectors configured'
e2e/specs/certificate-authorities/s9-hsm-connector-list-failure-shown-as-none.spec.ts:7
lines
infisical/frontend/src/pages/cert-manager/CertificateAuthoritiesPage/components/CreateCaWizard/CreateCaWizard.tsx:89
infisical/frontend/src/pages/cert-manager/CertificateAuthoritiesPage/components/CreateCaWizard/KeyValidityStep.tsx:174
infisical/frontend/src/hooks/api/hsmConnectors/queries.tsx:20
before the PRs testedno
An access token's Number of Uses stays at 0 however often it is used
s14-access-token-uses-never-counted-1
identity-authbehaviourloss: decision
high

When A machine uses a Token Auth access token that has a use limit.

Got The identity's access token list keeps showing 0 uses, however many times the token was used.

Want The list shows how many of its uses the token has spent.

steps
  1. As an org admin, open a machine identity and add Token Auth with an access token use limit of 5.
  2. Create an access token on its Token Auth method and copy it.
  3. Call any API (for example the identity read) twice with that access token as bearer; both succeed.
  4. Wait at least two minutes, reload the identity page and open Token Auth.
  5. Read the token's Number of Uses column.
s14-access-token-uses-never-counted
e2e/specs/identity-auth/s14-access-token-uses-never-counted.spec.ts:10
lines
infisical/backend/src/services/identity-access-token/identity-access-token-service.ts:510
infisical/backend/src/services/identity-access-token/identity-access-token-service.ts:351
infisical/backend/src/services/access-token-queue/access-token-queue.ts:62
infisical/backend/src/services/identity-access-token/identity-access-token-dal.ts:59
infisical/frontend/src/views/IdentityAuthMethods/helpers/IdentityTokenAuthTokensTable.tsx:157
before the PRs testedno
A machine token's use limit can be exceeded after its counter is lost from the cache
s14-lost-counter-reseed-overruns-use-limit-1
identity-authbehaviourloss: work
medium

When A machine token's use counter is lost from the cache (a cache flush) and two requests with the token arrive at the same moment.

Got The token works for one more request than its use limit allows.

Want The token stops working once its use limit is reached, even right after a cache flush.

steps
  1. Create a machine identity with Universal Auth and an access-token use limit of 3, and log in to get a token.
  2. Lose the token's remaining-use counter from Redis (the flush the product handles), simulated by deleting the identity-token-uses-remaining key.
  3. Make two requests with the token at the same moment so both read the missing counter before either re-seeds it (forced with an ACCESS EXCLUSIVE lock on identity_access_token_revocations that parks each request's revocation read after it has read the null counter).
  4. Release the lock, then read the counter and make further requests with the token until it is refused, counting the successful ones.
  5. Observe that the token authorized 4 requests in total and the counter read 2 right after the pair, instead of 3 and 1.
s14 lost-counter re-seed overruns use limit: a 3-use token whose counter is lost authorizes at most 3 requests
e2e/specs/identity-auth/s14-lost-counter-reseed-overruns-use-limit.spec.ts:54
lines
infisical/backend/src/services/identity-access-token/identity-access-token-service.ts:494
infisical/backend/src/services/identity-access-token/identity-access-token-service.ts:482
infisical/backend/src/services/identity-access-token/identity-access-token-service.ts:112
infisical/backend/src/keystore/keystore.ts:476
infisical/backend/src/services/identity-access-token/identity-access-token-service.ts:166
before the PRs testedno
Add Sync shows no destinations, with no error, when they fail to load
s4-destination-list-failure-shows-empty-1
uiloss: time
medium

When

Got

Want

steps
  1. Sign in as a project member who can manage secret syncs and open the project's Integrations page, Secret Syncs tab
  2. Make the request for the list of sync destinations fail with a server error
  3. Click Add Sync
  4. The "Choose a destination" dialog shows the search box and the "All providers" heading with no providers under it and no error
lines
infisical/frontend/src/components/secret-syncs/SecretSyncSelect.tsx:70
infisical/frontend/src/components/secret-syncs/SecretSyncSelect.tsx:97
infisical/frontend/src/components/secret-syncs/SecretSyncSelect.tsx:124
infisical/frontend/src/hooks/api/secretSyncs/queries.tsx:72
before the PRs testedno
The Variables list loads every service that uses a variable, with no limit
s15-variable-usage-lists-every-service-1
agent-vaultperformanceloss: time
medium

When Many services across your access bundles use the same variable.

Got Every visit to Variables, and every key check while adding a service, downloads the full list of those services.

Want The page loads a short summary per variable, whatever the project holds.

steps
  1. As an org admin, create an Agent Vault variable KEY
  2. Create two access bundles
  3. Add 78 services to each bundle, each with a bearer credential of {{KEY}}
  4. Look up KEY on the Variables list (one row: keys=KEY, limit 1)
s15-variable-usage-lists-every-service
e2e/specs/agent-vault/s15-variable-usage-lists-every-service.spec.ts:20
lines
infisical/backend/src/ee/services/agent-vault-variable/agent-vault-variable-dal.ts:171
infisical/backend/src/ee/services/agent-vault-variable/agent-vault-variable-service.ts:102
infisical/backend/src/ee/services/agent-vault-variable/agent-vault-variable-service.ts:61
infisical/frontend/src/pages/agent-vault/AgentVaultVariablesPage/components/UsedByBundleBadges.tsx:54
before the PRs testedno
The Variables list loads every listed variable's secret value though it never shows one
s15-variable-list-reads-sealed-values-1
agent-vaultperformanceloss: time
low

When You open the Agent Vault Variables list.

Got Behind the scenes every listed variable's secret value is loaded, then thrown away.

Want Only the keys and dates the list shows are loaded.

steps
  1. Sign in as an org admin of an organisation with Agent Vault
  2. Create 100 variables, each with an 8192-character value
  3. List the variables with a page size of 100 (search on their shared key prefix)
  4. Read the statement the store ran for the page in pg_stat_statements
s15-variable-list-reads-sealed-values
e2e/specs/agent-vault/s15-variable-list-reads-sealed-values.spec.ts:12
lines
infisical/backend/src/ee/services/agent-vault-variable/agent-vault-variable-dal.ts:37
infisical/backend/src/ee/services/agent-vault-variable/agent-vault-variable-dal.ts:67
infisical/backend/src/ee/services/agent-vault-variable/agent-vault-variable-service.ts:61
before the PRs testedno
Saving a typed storage hostname does nothing until Save is clicked again
s16-typed-storage-hostname-not-saved-by-save-1
server-adminbehaviourloss: time
medium

When You type a storage hostname into Allowed Storage Hostnames and click Save without pressing Enter first.

Got The hostname turns into a chip and Save lights up, but nothing is saved and no message appears.

Want One click on Save saves the hostname you typed.

steps
  1. Sign in as the instance super admin and open the Server Console's General page.
  2. In the Allowed Storage Hostnames card, type a hostname such as minio.example.com:9000 into Hostnames without pressing Enter.
  3. Click the card's Save once and wait a few seconds.
  4. Reload the page or read the instance config.
s16-typed-storage-hostname-not-saved-by-save
e2e/specs/server-admin/s16-typed-storage-hostname-not-saved-by-save.spec.ts:10
lines
infisical/frontend/src/components/v3/generic/TagsInput/TagsInput.tsx:187
infisical/frontend/src/components/v3/generic/TagsInput/TagsInput.tsx:85
infisical/frontend/src/pages/admin/GeneralPage/components/GeneralPageForm.tsx:99
infisical/frontend/src/pages/admin/GeneralPage/components/GeneralPageForm.tsx:480
#8485
reviewed by greptile, cubic, coderabbit
no
PAM template recording settings say no storage connections exist when they fail to load
s11-template-recording-connection-failure-shown-as-none-1
pamuiloss: time
medium

When You set a PAM template's session recordings to go to an S3 bucket and the connection list can't be loaded.

Got The connection picker says you have no AWS or S3-compatible connections and tells you to create one, with no error.

Want The picker tells you the connections couldn't be loaded.

steps
  1. Sign in as a user of an organisation with PAM on the enterprise plan.
  2. Create a PostgreSQL account template.
  3. Make the server fail the requests for the available AWS and S3-compatible connections.
  4. Open PAM > Templates and open the template's detail sheet on its general settings.
  5. Set Session Recording > Storage backend to S3 Bucket.
  6. Open the Connection picker.
s11-template-recording-connection-failure-shown-as-none
e2e/specs/pam/s11-template-recording-connection-failure-shown-as-none.spec.ts:6
lines
infisical/frontend/src/pages/pam/PamTemplatesPage/components/TemplateDetailSheet.tsx:284
infisical/frontend/src/pages/pam/PamTemplatesPage/components/TemplateDetailSheet.tsx:910
infisical/frontend/src/hooks/api/appConnections/queries.tsx:109
#8485
reviewed by greptile, cubic, coderabbit
no
A Windows account's recording override says no storage connections exist when they failed to load
s12-windows-recording-override-failed-connections-shown-as-none-1
pambehaviourloss: decision
medium

When You override a Windows account's recording storage while its connections can't be loaded.

Got The Connection picker says you have no storage connections and tells you to create one.

Want It tells you the connections could not be loaded.

steps
  1. Have a Windows PAM account in a folder, bound to a gateway, with its template.
  2. Open the account's detail sheet from its folder and go to the Advanced tab.
  3. While the organisation's app connection list fails to load, choose Override in Recording Settings.
  4. Open the Connection picker.
a Windows PAM account's recording override does not show a failed connection load as no connections
e2e/specs/pam/s12-windows-recording-override-failed-connections-shown-as-none.spec.ts:6
lines
infisical/frontend/src/pages/pam/PamAccountsPage/components/RecordingConnectionPicker.tsx:16
infisical/frontend/src/pages/pam/PamAccountsPage/components/RecordingConnectionPicker.tsx:34
infisical/frontend/src/pages/pam/PamAccountsPage/components/AccountDetailSheet.tsx:519
#8485
reviewed by greptile, cubic, coderabbit
no
A PAM template offers recording connections that saving then refuses
s12-template-recording-picker-offers-project-connection-1
pambehaviourloss: time
low

When You set a PAM template to record sessions to an S3 bucket and the PAM project has its own storage connection.

Got That connection is offered in the list, but saving the template with it is refused.

Want Only connections that can actually be saved are offered.

steps
  1. Create an S3-compatible storage connection at project level in the PAM project (the projectId the template sheet itself sends when listing connections).
  2. Create a PostgreSQL PAM account template.
  3. Open PAM > Account Templates, open the template's general settings.
  4. Set Session Recording storage backend to S3 Bucket and open the Connection picker.
  5. Choose the project-level connection and save.
a PAM template's recording connection picker does not offer a project-level connection
e2e/specs/pam/s12-template-recording-picker-offers-project-connection.spec.ts:8
lines
infisical/frontend/src/pages/pam/PamTemplatesPage/components/TemplateDetailSheet.tsx:284
infisical/frontend/src/pages/pam/PamTemplatesPage/components/TemplateDetailSheet.tsx:916
infisical/backend/src/ee/services/pam/pam-validators.ts:115
infisical/frontend/src/pages/pam/PamAccountsPage/components/RecordingConnectionPicker.tsx:18
#8485
reviewed by greptile, cubic, coderabbit
no
A storage hostname saved from one admin tab is dropped by a save from another
s16-stale-admin-tab-drops-saved-storage-hostname-1
server-adminbehaviourloss: work
low

When Two admin console tabs are open and each adds a storage hostname and saves.

Got Both say the hostnames were saved, but the first tab's hostname is gone from the list.

Want Both hostnames are kept, or the second save is stopped as out of date.

steps
  1. Sign in as a server admin and open the Server Console's General page in two tabs
  2. In the first tab, add a hostname such as host-c.example.com:9000 to Allowed Storage Hostnames and click Save; the success notification shows
  3. In the second tab, without reloading, add host-d.example.com:9000 and click Save; the success notification shows
  4. Reload either tab and read the Allowed Storage Hostnames list
s16-stale-admin-tab-drops-saved-storage-hostname
e2e/specs/server-admin/s16-stale-admin-tab-drops-saved-storage-hostname.spec.ts:36
lines
infisical/frontend/src/pages/admin/GeneralPage/components/GeneralPageForm.tsx:458
infisical/frontend/src/pages/admin/GeneralPage/components/GeneralPageForm.tsx:466
infisical/frontend/src/pages/admin/GeneralPage/components/GeneralPageForm.tsx:532
infisical/frontend/src/context/ServerConfigContext/ServerConfigContext.tsx:9
infisical/backend/src/server/routes/v1/admin-router.ts:188
infisical/backend/src/services/super-admin/super-admin-service.ts:496
#8485
reviewed by greptile, cubic, coderabbit
no
Saving a PAM template can silently turn off another admin's approval requirement
s13-stale-template-sheet-keeps-newer-approval-1
pambehaviourloss: work
low

When You save a PAM account template's settings in a sheet you opened before another admin turned on Require Approval for it.

Got Your save quietly turns Require Approval back off.

Want Only your change is saved, or you are told the template changed since you opened it.

steps
  1. As an org admin, create a PostgreSQL PAM account template with no policies.
  2. Open PAM > Account Templates and open the template's detail sheet on the General tab.
  3. As another admin (here: the API with the same org's token), turn on the template's Require Approval policy.
  4. Back in the still-open sheet, change only Min length from 32 to 40 and click Save Changes.
  5. Read the template back.
s13-stale-template-sheet-keeps-newer-approval
e2e/specs/pam/s13-stale-template-sheet-keeps-newer-approval.spec.ts:5
lines
infisical/frontend/src/pages/pam/PamTemplatesPage/components/TemplateDetailSheet.tsx:425
infisical/frontend/src/pages/pam/PamTemplatesPage/components/TemplateDetailSheet.tsx:483
infisical/backend/src/ee/services/pam-account-template/pam-account-template-service.ts:241
infisical/backend/src/ee/services/pam-account-template/pam-account-template-service.ts:242
infisical/backend/src/ee/services/pam-account-template/pam-account-template-service.ts:251
before the PRs testedno
PAM Accounts cuts off its page title and folder names on a phone-width screen
s13-accounts-header-off-screen-on-phone-1
pamuiloss: time
low

When PAM Accounts is opened on a phone-width screen.

Got The page sits shifted sideways: the Accounts title and the start of every folder name are cut off, and swiping does not bring them back.

Want The whole page fits the screen, title, folder names and Add Account included.

steps
  1. Open an organisation with PAM on a screen 390 or 320 pixels wide
  2. Go to PAM > Accounts
  3. Look at the top of the page, then try to swipe sideways
PAM Accounts' title and Add Account are reachable at ${size.name} width (${size.width}px)
e2e/specs/pam/s13-accounts-header-off-screen-on-phone.spec.ts:7
lines
infisical/frontend/src/pages/pam/PamAccountsPage/PamAccountsPage.tsx:253
infisical/frontend/src/pages/pam/PamAccountsPage/PamAccountsPage.tsx:254
before the PRs testedno
On a phone-width screen an application's certificate search field is squeezed to nothing and covered by the filter button
s18-certificate-search-collapsed-at-phone-width-1
pki-syncs-alertsuiloss: time
low

When You open an application's certificates on a phone-width screen.

Got The search field is squeezed to a sliver and the filter button sits on top of it, so you cannot tap into it to search.

Want The search field stays wide enough to type in.

steps
  1. Sign in and open Certificate Manager in an organization with an active Certificate Manager project
  2. Create or open an application, or open the project's Inventory
  3. View the Certificate Inventory on a phone-width screen (390 px or 320 px wide)
  4. Try to tap the search field above the certificate list
lines
infisical/frontend/src/pages/cert-manager/CertificatesPage/components/CertificatesTable.tsx:733
infisical/frontend/src/pages/cert-manager/CertificatesPage/components/CertificatesTable.tsx:734
infisical/frontend/src/pages/cert-manager/CertificatesPage/components/CertificatesTable.tsx:803
before the PRs testedno
A revocation reported as failed still revokes the certificate but leaves it off the revocation list
s14-failed-revocation-left-off-crl-1
certificatesbehaviourloss: decision
medium

When You revoke a certificate while the CA's signing key in AWS KMS cannot sign.

Got You are told the revocation failed, but the certificate shows as revoked, it is missing from the CA's revocation list, and revoking it again fails.

Want The revocation either fully fails so you can retry, or the certificate is on the revocation list.

steps
  1. Create an AWS connection and an internal root CA whose key source is AWS KMS, a certificate policy and an API-enrollment certificate profile on that CA.
  2. Issue a certificate from the profile and revoke it: the CA's revocation list lists its serial (control).
  3. Issue a second certificate from the profile.
  4. Make the CA's KMS signing key unable to sign (disable it in KMS).
  5. Revoke the second certificate: the answer is an error, 'Failed to sign with ... in AWS KMS: The AWS KMS key is disabled.'
  6. Re-enable the key, then read the certificate, read the CA's revocation list, and revoke the certificate again.
a certificate whose revocation failed while the CRL was re-signed is still revocable or listed on the CRL
e2e/specs/certificates/s14-failed-revocation-left-off-crl.spec.ts:55
lines
infisical/backend/src/services/certificate/certificate-service.ts:757
infisical/backend/src/services/certificate/certificate-service.ts:800
infisical/backend/src/services/certificate/certificate-service.ts:714
infisical/backend/src/services/certificate-authority/certificate-authority-fns.ts:600
infisical/backend/src/services/certificate-authority/certificate-authority-queue.ts:160
before the PRs testedno
Insights header hides the last sent report with no error when reports fail to load
s14-failed-org-report-list-hides-last-report-1
secret-insightsuiloss: decision
medium

When Your organisation's report history can't be loaded on the Insights page.

Got The "Last report sent" date disappears and nothing says anything went wrong.

Want The page keeps the date or tells you the reports couldn't be loaded.

steps
  1. Sign in as an organisation admin whose organisation has a sent insights report.
  2. Open the organisation's Insights page (Secrets Management → Insights): the header reads "Last report sent <date>" beside Generate Report.
  3. Make the organisation's report list fail to load with a server error (its first try and the app's retry).
  4. Reload the Insights page.
s14-failed-org-report-list-hides-last-report
e2e/specs/secret-insights/s14-failed-org-report-list-hides-last-report.spec.ts:16
lines
infisical/frontend/src/pages/organization/SecretInsightsPage/SecretInsightsPage.tsx:108
infisical/frontend/src/pages/organization/SecretInsightsPage/SecretInsightsPage.tsx:115
infisical/frontend/src/pages/organization/SecretInsightsPage/SecretInsightsPage.tsx:150
infisical/frontend/src/pages/organization/SecretInsightsPage/SecretInsightsPage.tsx:165
before the PRs testedno
A variable still shows as unused after its delete is refused for being in use
s1-refused-delete-leaves-variable-unused-1
agent-vault-variablesbehaviourloss: time
medium

When Someone starts using a variable in a service while you are deleting it.

Got Deleting is refused and a toast says why, but the variable still reads Unused and the dialog still offers Delete.

Want The page updates to show which service uses the variable and stops offering Delete.

steps
  1. As an org admin, add an Agent Vault variable that no service uses; its row on the Variables page reads Unused.
  2. Open the variable's Delete dialog: it reads 'Its value is deleted with it. This cannot be undone.' and Delete Variable is enabled.
  3. Meanwhile (another admin or another tab) add a service to an access bundle whose bearer credential references {{KEY}} of that variable.
  4. Click Delete Variable: an error toast says the variable is used by the service 'svc-uses-var' and the deletion is refused.
  5. The dialog keeps saying its value is deleted with it and keeps Delete Variable enabled; close it with Cancel.
  6. Look at the variable's row.
s1-refused-delete-leaves-variable-unused
e2e/specs/agent-vault-variables/s1-refused-delete-leaves-variable-unused.spec.ts:10
lines
infisical/frontend/src/hooks/api/agentVault/mutations.tsx:443
infisical/frontend/src/hooks/api/agentVault/mutations.tsx:429
infisical/frontend/src/pages/agent-vault/AgentVaultVariablesPage/components/DeleteVariableDialog.tsx:30
infisical/frontend/src/pages/agent-vault/AgentVaultVariablesPage/components/DeleteVariableDialog.tsx:81
infisical/backend/src/ee/services/agent-vault-variable/agent-vault-variable-service.ts:200
before the PRs testedno
Two same-name KMS-backed CAs created at once leave an extra key in AWS
s10-same-name-kms-cas-leave-second-key-1
certificate-authoritiesbehaviourloss: nothing
low

When Two people (or a retrying script) create an AWS KMS-backed CA with the same name at the same moment.

Got One CA is created and the other refused, but a second signing key is created in your AWS account and left pending deletion.

Want The duplicate is refused before any key is made in AWS, as it is when the creates come one after the other.

steps
  1. Create an AWS app connection in a Certificate Management project.
  2. Create an AWS KMS-backed root CA named N, then create N again: the second is refused and AWS KMS holds one key described 'Infisical certificate authority signing key: N'.
  3. Send two creates of an AWS KMS-backed root CA with a new name N2 at the same moment.
  4. List the AWS KMS keys whose description is 'Infisical certificate authority signing key: N2'.
two AWS KMS-backed CAs created together under one name leave one CA and one KMS signing key
e2e/specs/certificate-authorities/s10-same-name-kms-cas-leave-second-key.spec.ts:50
lines
infisical/backend/src/services/certificate-authority/internal/internal-certificate-authority-service.ts:439
infisical/backend/src/services/certificate-authority/internal/internal-certificate-authority-service.ts:447
infisical/backend/src/services/certificate-authority/internal/internal-certificate-authority-service.ts:545
infisical/backend/src/services/certificate-authority/internal/internal-certificate-authority-service.ts:633
infisical/backend/src/services/certificate-authority/ca-key-source/aws-kms-ca-key-source.ts:138
#8637
reviewed by greptile, cubic, coderabbit
no
On a phone-width screen the project header's organization link and switchers cannot be tapped
s9-header-switchers-unreachable-at-narrow-width-1
secret-syncsuiloss: time
low

When An organization admin opens a project page on a phone-width screen.

Got The organization name link disappears, and at 320px a tap on either switcher hits the block beside it or opens the Invite Users page.

Want The organization link and both switchers stay tappable at phone widths.

steps
  1. Sign in as an organization admin and open any project's Integrations page, Secret Syncs tab
  2. Narrow the window to 390px wide: the organization name link in the header has disappeared (0px wide)
  3. Narrow it to 320px wide and tap the organization switcher chevron: the tap lands on the project-type block beside it
  4. Tap the project switcher chevron: the tap lands on the Invite Users link and opens the members page
lines
infisical/frontend/src/layouts/OrganizationLayout/components/NavBar/Navbar.tsx:408
infisical/frontend/src/layouts/OrganizationLayout/components/NavBar/Navbar.tsx:334
infisical/frontend/src/layouts/OrganizationLayout/components/NavBar/Navbar.tsx:467
infisical/frontend/src/layouts/ProjectLayout/components/TypeSelect/TypeSelect.tsx:123
infisical/frontend/src/layouts/ProjectLayout/components/ProjectSelect/ProjectSelect.tsx:161
infisical/frontend/src/layouts/OrganizationLayout/components/NavBar/Navbar.tsx:650
before the PRs testedno
Variable suggestions in a service's token field load forever when the list fails
s16-variable-suggestions-stuck-loading-on-failed-list-1
agent-vaultbehaviourloss: time
medium

When You type {{ in a service's token field and the variable list can't be loaded.

Got The suggestions keep saying "Loading variables..." and never finish or explain.

Want The suggestions say the variables could not be loaded.

steps
  1. Sign in as an org admin with Agent Vault available, and add a variable on Agent Vault > Variables.
  2. Open an access bundle and start Add Service; pick the OpenAI template, give it a name and press Continue.
  3. Make the variable list request fail with a server error.
  4. Click the token field and type "{{AGENT".
s16-variable-suggestions-stuck-loading-on-failed-list
e2e/specs/agent-vault/s16-variable-suggestions-stuck-loading-on-failed-list.spec.ts:10
lines
infisical/frontend/src/components/agent-vault/service-sheet/VariableSuggestions.tsx:182
infisical/frontend/src/components/agent-vault/service-sheet/VariableSuggestions.tsx:44
infisical/frontend/src/components/agent-vault/service-sheet/VariableSuggestions.tsx:312
before the PRs testedno
Agent Vault Variables page says there are no variables when the list fails to load
s15-variable-list-failure-shown-as-no-variables-1
agent-vaultbehaviourloss: decision
medium

When You open Agent Vault Variables and the list can't be loaded.

Got The page says there are no variables yet and offers to add one, though your organisation has some.

Want The page says the variables could not be loaded.

steps
  1. Sign in as an org admin with Agent Vault available.
  2. Add a variable on Agent Vault > Variables.
  3. Make the variable list request fail with a server error (500), including the automatic retry.
  4. Open Agent Vault > Variables.
lines
infisical/frontend/src/pages/agent-vault/AgentVaultVariablesPage/components/VariablesCard.tsx:229
infisical/frontend/src/pages/agent-vault/AgentVaultVariablesPage/components/VariablesCard.tsx:245
infisical/frontend/src/pages/agent-vault/AgentVaultVariablesPage/components/VariablesCard.tsx:362
infisical/frontend/src/hooks/api/agentVault/queries.tsx:545
before the PRs testedno
Closing Add Variable throws away the typed variable key without asking
s17-variable-key-draft-wiped-on-escape-1
agent-vaultuiloss: time
low

When You type a key in Add Variable and close the form without saving, even by pressing Escape.

Got The form closes at once without asking, and when you open it again the key you typed is gone.

Want The product asks before throwing your variable away, or keeps what you typed.

steps
  1. Sign in as an organization admin and open Agent Vault, Variables
  2. Click Add Variable and type a variable key (and a value)
  3. Close the form with Escape or Cancel (the X and a click outside do the same)
  4. Click Add Variable again
lines
infisical/frontend/src/components/agent-vault/VariableFormDialog.tsx:305
infisical/frontend/src/components/agent-vault/VariableFormDialog.tsx:282
infisical/frontend/src/components/agent-vault/VariableFormDialog.tsx:313
before the PRs testedno