aevatarAI · AbigailDeng · Jul 31, 2026 · Jul 31, 2026
@@ -0,0 +1,176 @@
+# Frontend Test Suite Risk Audit
+
+## Scope and Method
+
+This audit applies the risk-based rules in `docs/testing-policy.md` to the
+current `dev` inventory of `src/**/*.test.{ts,tsx}`. Commit `4a7d49f3c` carries
+the policy and is an ancestor of the audited branch.
+
+The inventory contains 138 test files. Exactly 38 files exceed 300 lines at
+the audited commit. The review considered test names, fixture boundaries,
+repeated setup, mock and collaborator-call assertions, observable assertions,
+and overlap with adjacent test files. Line count and mock count were used only
+as review triggers, not as deletion criteria.
+
+This pull request performs one bounded migration in the Studio GAgent catalog
+discovery and published-revision identity domain. The old route-level case
+`loads discovered GAgent types and the published service revision catalog`
+asserted only that two mocked collaborators were called. It would fail after an
+internal data-loading refactor even when the product behavior was unchanged,
+while still passing if the returned catalog data never reached the operator.
+
+The migration replaces that weak evidence with three observable contracts:
+
+- The Studio route now renders the real `StudioGAgentBuildPanel` while mocking
+  only the heavy script-editor boundary. A discovered `Orders Assistant` kind
+  must reach the GAgent selector and the real definition fields must render.
+- A focused component integration case selects `Billing Assistant` through the
+  real selector, continues to Bind, and observes that the handoff retains
+  `Tests.BillingGAgent` rather than the previously selected kind.
+- The initial Team member rail must render the published workflow name and the
+  exact service/revision identity for `billing-api · rev-billing-7`, rather
+  than merely proving that a revision API was invoked.
+
+The workflow, Script, and GAgent service-revision fixtures are consolidated in
+`tests/studioServiceCatalogFixtures.ts` so future behavior-domain extraction
+does not duplicate the oversized route fixture block. No production code
+changes are made.
+
+## Priority Behavior and Risk Inventory
+
+### `src/pages/studio/index.test.tsx`
+
+| Behavior domain | Concrete risk | Highest-value evidence | Review outcome |
+| --- | --- | --- | --- |
+| Route and identity canonicalization | A service, member, workflow, or legacy query identity is sent to the wrong authority or survives in a canonical URL | Component or route integration | Retain the distinct initial-load, route-change, and missing-target cases; replace the duplicate bootstrap call-only case with visible service/revision identity evidence |
+| Team member inventory and entry selection | Cross-Team members leak into the rail or the wrong member becomes the Team entry | Component or route integration | Retain; assertions act through the rendered rail and entry action |
+| Workflow, Script, and GAgent authoring handoffs | The selected implementation or unsaved draft is lost across Build and Bind | Component or route integration | Split later by implementation behavior when shared setup can be extracted without recreating the mock graph |
+| Bind observation and published contract selection | A pending or stale binding is presented as callable, or the scope default replaces the selected member | Component or route integration | Retain; these are high-risk identity and eventual-consistency rules |
+| Invoke and Observe handoffs | Runs are started or stopped for the wrong member, service, or run | Component or route integration | Retain; adapter request-shape checks are product identity contracts here |
+| Auth recovery | Studio loops through login or loads protected state before session recovery settles | Component or route integration | Retain locally; real redirect and provider acceptance remain a browser smoke gap |
+
+The file contains several independent behavior domains and oversized shared
+fixtures. Its outcome is **split by behavior domain**, but only after a shared
+route render kit can preserve realistic internal collaboration. Mechanical
+file splitting would duplicate more than 3,000 lines of setup and increase the
+maintenance risk this audit is intended to reduce.
+
+### `src/pages/team-member-workflow-studio/index.test.tsx`
+
+| Behavior domain | Concrete risk | Highest-value evidence | Review outcome |
+| --- | --- | --- | --- |
+| New member and draft materialization | A retry creates duplicate members or links the wrong workflow draft | Component or route integration | Retain; identity fixtures are deliberately distinct |
+| Route draft recovery | Member, workflow, and published-service identities are substituted for one another | Component or route integration | Retain and split from editor interaction coverage |
+| Graph and YAML editing | Graph, YAML, and dirty-state buffers diverge or stale async work overwrites newer edits | Component integration, with pure document rules in `shared/studio/document.test.ts` | Retain integration cases; move only deterministic transformation duplication to the helper suite |
+| Guided node configuration | Typed and raw parameter editors write different runtime shapes | Component integration | Retain; it protects the editor-to-document boundary |
+| Draft execution and streaming logs | Unsaved input/files are omitted or stream frames produce the wrong visible result | Component integration | Retain; real SSE transport and browser streaming remain a browser smoke gap |
+| Publish and rebind observation | Pending, failed, stale, or already-published facts expose the wrong action | Component or route integration | Retain and split as a publish-observation behavior domain |
+
+The outcome is **split by behavior domain**. The most valuable boundaries are
+route identity, editor synchronization, draft execution, and publish
+observation. Pointer resizing and native leave-prompt behavior cannot be fully
+proven by jsdom and remain browser gaps.
+
+### `src/pages/teams/detail.test.tsx`
+
+| Behavior domain | Concrete risk | Highest-value evidence | Review outcome |
+| --- | --- | --- | --- |
+| Team shell and navigation | Scope or Team context is lost across tabs and breadcrumbs | Component or route integration | Retain |
+| Roster, entry member, and deletion | Accepted commands are shown as completed before the read model confirms them | Component or route integration | Retain; these cases protect destructive and eventual-consistency behavior |
+| Team Test | A test invokes the selected row rather than the authoritative entry member, or runs while entry state is stale | Component or route integration | Retain and split as an independent Team Test domain |
+| Member workflow handoffs | Route hints override authoritative member, workflow, or service identities | Component or route integration | Retain |
+| Member automation handoffs | Query hints become owner identity or cross-Team members are queried | Component or route integration | Retain |
+| Rename, archive, and projection recovery | An accepted update is reverted by a stale refresh or a transient 404 becomes data loss | Component or route integration | Retain |
+
+The outcome is **split by behavior domain**. The responsive row-action case is
+useful as a layout regression signal, but actual clipping and Team Test
+streaming across deployed services remain browser smoke gaps.
+
+### `src/shared/studio/api.test.ts`
+
+| Behavior domain | Concrete risk | Highest-value evidence | Review outcome |
+| --- | --- | --- | --- |
+| Authenticated host requests and errors | Bearer auth, problem details, or compact HTML errors are adapted incorrectly | Module or adapter integration | Retain |
+| Workflow draft and published fallback | Scope is omitted or a missing draft hides a committed workflow | Module or adapter integration | Retain and split by workflow persistence behavior |
+| Binding contracts | Member and scope endpoints or stable workflow identities are confused | Module or adapter integration | Retain |
+| Team and member commands | Typed identities, explicit nulls, or accepted command responses are lost during adaptation | Module or adapter integration | Retain and split by authority resource |
+| Workflow board mapping | Nullable totals and current-state snapshot fields are decoded incorrectly | Module or adapter integration | Retain |
+
+The outcome is **split by adapter behavior domain**. Fetch is the external
+boundary; internal decoders and request builders remain real. The suite should
+not be converted into route tests, and its request-shape assertions are
+observable transport contracts rather than incidental call forwarding.
+
+### `src/pages/studio/components/StudioBuildPanels.test.tsx`
+
+| Behavior domain | Concrete risk | Highest-value evidence | Review outcome |
+| --- | --- | --- | --- |
+| Script authoring and save observation | Unsaved packages are lost or pending/rejected saves enable Bind | Component integration | Retain and split as Script authoring |
+| Workflow prompt and node editing | User instructions are written to the wrong runtime parameter or omitted from save/run | Component integration | Retain; pure mapping overlap belongs in document-helper tests |
+| Diagnostics and mutation locks | Invalid JSON or in-flight mutations allow Apply, Save, or Run | Component integration | Retain |
+| Draft-run output | Metadata or intermediate frames replace the final product output | Component integration | Retain |
+| GAgent discovery and handoff | A discovered kind never reaches Build, or selecting a different kind still hands the stale kind to Bind | Route plus component integration | Cover discovery through the real route-owned panel and selection through the real selector |
+| GAgent draft-run recovery | Provider failure escapes Build without a retry path | Component integration | Retain |
+
+The outcome is **split by behavior domain**. The legacy-surface deletion guard
+is a temporary architecture assertion and should be removed with that surface,
+not expanded. Real provider routing remains a browser or deployed smoke gap.
+
+## Review Outcomes for Every File Above 300 Lines
+
+| File | Lines | Recorded outcome and rationale |
+| --- | ---: | --- |
+| `src/pages/studio/index.test.tsx` | 8,617 | **Split by behavior domain.** Replace duplicate bootstrap call-only coverage with observable catalog/revision behavior now; later isolate route identity, inventory, authoring, Bind, Invoke, and Observe around a shared render kit. |
+| `src/pages/team-member-workflow-studio/index.test.tsx` | 7,559 | **Split by behavior domain.** Route identity, editor synchronization, draft execution, and publish observation have independent fixtures and risks. |
+| `src/pages/teams/detail.test.tsx` | 3,219 | **Split by behavior domain.** Roster mutation, Team Test, handoffs, and Team administration are distinct route behaviors. |
+| `src/shared/studio/api.test.ts` | 2,997 | **Split by behavior domain.** Keep module/adapter integration and separate workflow, binding, Team, member, and board contracts. |
+| `src/pages/chat/index.test.tsx` | 1,991 | **Split by behavior domain.** Conversation recovery, streaming, history, and Studio handoff risks should not share one route fixture indefinitely. |
+| `src/pages/studio/components/StudioMemberInvokePanel.test.tsx` | 1,867 | **Split by behavior domain.** Separate chat streaming, structured invoke, attachments, run history, and failure recovery. |
+| `src/pages/studio/components/StudioBuildPanels.test.tsx` | 1,796 | **Split by behavior domain.** Separate Script authoring, Workflow editing, validation, and draft-run output. |
+| `src/pages/settings/index.test.tsx` | 1,246 | **Reduce duplicated setup, then split.** Exact-service observation races are valuable but share large catalog and async fixtures. |
+| `src/pages/studio/components/bind/StudioMemberBindPanel.test.tsx` | 1,227 | **Retain with rationale.** Thirteen cases cover distinct contract states; extract fixture builders rather than split mechanically. |
+| `src/pages/scopes/invoke.test.tsx` | 1,084 | **Reduce repeated fixtures.** Eight route-integration cases cover distinct chat, typed invoke, reset, and runtime handoff behavior. |
+| `src/pages/MissionWall/index.test.tsx` | 1,067 | **Retain with rationale.** Snapshot freshness, focus stability, and authoritative workflow-board rendering are distinct fullscreen behaviors. |
+| `src/pages/teams/home.test.tsx` | 1,052 | **Split by behavior domain.** Team inventory, entry-member runtime sampling, archive filtering, and projection recovery are independent. |
+| `src/pages/runs/index.test.tsx` | 992 | **Split by behavior domain.** Launch setup, run restoration, streaming retry, and human interaction are independent route risks. |
+| `src/locales/hardcodedCopyAudit.test.ts` | 917 | **Retain with rationale.** Length comes from deterministic source fixtures for one repository-wide i18n invariant, with no mock graph. |
+| `src/shared/api/scheduledDispatchApi.test.ts` | 888 | **Split by behavior domain.** Listing, mutation encoding, retry, validation, and action routes are separate adapter contracts. |
+| `src/pages/settings/nyxIdRelayLlm.test.ts` | 866 | **Reduce duplicated permutations.** Preserve distinct exact-identity, stale observation, race, timeout, and fail-closed risks. |
+| `src/pages/teams/tabs/TeamAutomationsTab.test.tsx` | 840 | **Split by behavior domain.** Admission review, create/update, credential recovery, and revocation observation are independent workflows. |
+| `src/shared/api/runtimeRunsApi.test.ts` | 771 | **Split by endpoint behavior.** Keep request-shape coverage because member, Team, service, draft, and typed invoke routes are public contracts. |
+| `src/pages/chat/chatApi.test.ts` | 692 | **Retain with rationale.** SSE parsing, continuation recovery, cancellation, and structured frame extraction are distinct adapter risks. |
+| `src/shared/api/scopeRuntimeApi.test.ts` | 689 | **Retain with rationale.** Eleven concise adapter cases cover distinct service, binding, endpoint, run, and retirement contracts. |
+| `src/shared/api/teamAutomationApi.test.ts` | 681 | **Split by behavior domain.** Authorization decoding, fail-closed admission, owner-scoped schedules, and revocation are independent. |
+| `src/pages/Deployments/index.test.tsx` | 662 | **Split by behavior domain.** Inventory states, handoff routing, rollout control, and evidence transitions are independent route risks. |
+| `src/pages/gagents/index.test.tsx` | 626 | **Reduce repeated setup.** Six high-value route cases protect discovery, draft run, binding replacement, activation, and retirement. |
+| `src/pages/studio/components/StudioFilesPage.test.tsx` | 612 | **Split by behavior domain.** Catalog browsing, chat-history deletion, and Chrono file editing have independent boundaries. |
+| `src/shared/studio/document.test.ts` | 594 | **Retain with rationale.** Pure document transformations are correctly placed at the unit layer and protect distinct graph invariants. |
+| `src/shared/auth/client.test.ts` | 547 | **Retain with rationale.** OAuth initiation, callback, service-access review, refresh, and stale-storage races are distinct adapter behaviors. |
+| `src/pages/runtime-published-runs/index.test.tsx` | 503 | **Retain with rationale.** Seven cases cover loading, canonical routing, schedule filtering, materialization, and back navigation. |
+| `src/pages/team-member-invoke/index.test.tsx` | 430 | **Retain with rationale.** Seven route cases enforce distinct member/workflow/service identity and binding eligibility rules. |
+| `src/shared/studio/nodeConfigFields.structured.test.ts` | 411 | **Retain with rationale.** Pure structured-editor mappings and validation belong at the unit layer and use no mock graph. |
+| `src/pages/Deployments/releaseEvidence.test.ts` | 409 | **Retain with rationale.** Pure evidence state transitions are correctly placed and cover distinct rollout/deactivate/rollback states. |
+| `src/shared/api/runtimeGAgentApi.test.ts` | 388 | **Retain with rationale.** Nine adapter cases cover discovery, saved actors, draft runs, binding, activation, and retirement contracts. |
+| `src/adminObservatoryGraph.test.ts` | 380 | **Retain with rationale.** Two dense pure graph cases share one topology fixture and protect edge preservation and interaction layout. |
+| `src/pages/governance/components/GovernanceInspectorDrawer.test.tsx` | 367 | **Reduce repeated mock setup.** Six component cases protect distinct create, update, retired-state, and exposure behavior. |
+| `src/shared/navigation/teamRoutes.test.ts` | 356 | **Retain with rationale.** Pure route builders/parsers encode high-risk identity separation across distinct canonical routes. |
+| `src/shared/graphs/GraphCanvas.test.tsx` | 322 | **Retain with rationale.** Seven component cases cover deletion ownership, labels, viewport, positions, and connection handles. |
+| `src/pages/studio/components/StudioExecutionPage.test.tsx` | 321 | **Retain with rationale.** Six component cases cover runtime context, approval, signals, script downgrade, and empty state. |
+| `src/pages/runs/components/RunsLaunchRail.test.tsx` | 309 | **Reduce oversized fixtures.** Four observable layout and disclosure cases do not need the current mock volume. |
+| `src/pages/chat/chatHistoryApi.test.ts` | 308 | **Retain with rationale.** Eight adapter cases cover pagination, recovery identity, decoding, empty data, malformed data, deletion, and errors. |
+
+## Residual Coverage Gaps
+
+- OAuth redirect origin, provider callback acceptance, and cross-origin session
+  recovery require browser smoke coverage.
+- Native browser history, leave prompts, pointer resizing, responsive clipping,
+  and focus behavior are only approximated by jsdom.
+- Real SSE framing, cancellation, multipart upload, and deployed Team/member
+  routing require a focused browser or deployed integration check.
+- Backend projection timing and binding materialization are represented with
+  deterministic controlled responses; the suite does not prove production
+  timing or infrastructure wiring.
+
+These gaps are recorded rather than replaced with additional mock-heavy unit
+tests. No browser, end-to-end, integration, smoke, or full frontend suite is
+authorized or run by this audit.