Problem: No tests verify that ambient mode actually triggers skills. We need integration test infrastructure to validate that the ambient-router correctly activates skills based on prompt intent.
Solution:
- Create integration test infrastructure using
claude -p with output parsing
- Separate vitest config with 60-second timeout for integration tests (vs 5s for unit tests)
- Test infrastructure validates:
- Ambient mode classification (QUICK/STANDARD/ESCALATE)
- Correct skills loaded for each classification
- Skill content injected into context
- Skip logic working (slash commands, confirmations)
Test Coverage:
- Various prompt intents (refactor, test, debug, document, etc.)
- Edge cases (empty prompts, slash commands, very short responses)
- Skill loading correctness for each depth tier
Acceptance Criteria:
- Integration tests pass with 100% success
- Vitest config supports 60s timeout for integration tests
- All ambient-router classification paths tested
- Tests runnable via
npm test:integration
Depends on: #79 (ambient session injection)
Related to: #78
Problem: No tests verify that ambient mode actually triggers skills. We need integration test infrastructure to validate that the ambient-router correctly activates skills based on prompt intent.
Solution:
claude -pwith output parsingTest Coverage:
Acceptance Criteria:
npm test:integrationDepends on: #79 (ambient session injection)
Related to: #78