/retrospective — Session Analysis

You are executing the

/retrospective

command. This is script-driven — it runs the session evaluation pipeline.

What This Does

Analyzes the completed coding session to evaluate quality, identify improvements, and suggest harness enhancements.

Process

1. Run the evaluation script:

node .claude/hooks/scripts/evaluate-session.js

This script analyzes:

• Files created/modified during the session
• Test results and coverage changes
• Build success/failure history
• Hook trigger patterns (which rules fired, which were ignored)
• Time spent in different phases (planning, coding, testing, debugging)

2. Generate session report:

## Session Retrospective

### Work Summary
- Files changed: 12
- Tests added: 8
- Tests passing: 47/47
- Build status: Success

### Quality Metrics
- Type safety: 100% (no new `any`)
- Test coverage delta: +3.2%
- Bundle size delta: +8KB

### Patterns Observed
- [GOOD] Consistent use of error boundaries
- [GOOD] All new components have tests
- [IMPROVE] 3 files missing JSDoc on public API
- [IMPROVE] 2 effects missing cleanup

### Harness Feedback
- Rule `common/state-management.md` triggered 5 times → Well calibrated
- Rule `expo/patterns.md` never triggered → May need broader globs
- Hook `lint-staged.cjs` caught 2 issues → Working as intended
- Suggestion: Add rule for consistent error message format

### Recommendations
1. Add error message formatting rule
2. Review expo/patterns.md glob coverage
3. Consider adding pre-commit test hook

3. Act on recommendations: Review each recommendation and decide whether to implement it now or add to backlog.

When to Use

• At the end of a significant coding session
• After completing a feature or milestone
• When the harness feels miscalibrated (too many or too few rule triggers)
• Periodically for continuous improvement