Summary

Type: SPIKE — audit only; fixes spawn as follow-up items.

#2336 Area 8 ("Test-layer hardening") identified global mock shields that hid production bugs — e.g., runPreparedReply mock in src/auto-reply/reply/get-reply.test-mocks.ts silently replaced a real function with vi.fn(async () => undefined), preventing integration tests from exercising the real call chain. Analogous patterns may exist in UI tests: fixtures referencing gutted features that silently validate code paths no longer in production.

Approach

1. Grep ui/src/**/*.test.ts for gutted-feature references:
   • skills, Skills, agentSkills
   • plugin, Plugin, pluginId
   • thinkingLevel, chatThinkingLevel, thinkLevel
   • autoAllowSkills
2. For each hit, classify:
   • DEAD FIXTURE: test targets gutted functionality (feature is gone, test still runs) — candidate for removal
   • MOCK SHIELD: test fixture masks a real call that would otherwise fail (analogous to gut(auto-reply): finish thinking-level and model-selection sweep — follow-up to #2335 #2336 Area 8 finding) — high-priority cleanup
   • LEGITIMATE: tests live pass-through feature (e.g., <think> tag stripping) — KEEP, not a remnant

Deliverable

Classified inventory. For MOCK SHIELD items, open a high-priority follow-up issue with specific integration tests to add (per #2336 Area 8 pattern). For DEAD FIXTURE items, open a cleanup issue or consolidate into one test-cleanup PR.

AC

• Classified inventory (table of all hits with file:line + classification)
• Follow-up issues opened per category (DEAD FIXTURE sweep + MOCK SHIELD per-item)

Typed exception

SPIKE — AC is the audit inventory, not test modifications.

Context

Post-#2336 UI remnants audit. This spike extends the test-layer hardening theme from #2336 Area 8 into the UI test suite specifically.