AgentSkillsCN

Milestone Review

里程碑评审

SKILL.md

/milestone-review — End-of-Milestone Validation

Run this skill at the end of each milestone to validate completion and check for drift.

Usage

code
/milestone-review <milestone-number>

Example: /milestone-review 2

Workflow

1. Read the Milestone Spec

Read docs/milestones/M<N>.md for the completed milestone. Note the use cases, files, endpoints, and Definition of Done checklist.

2. Walk Through Definition of Done

For each item in the DoD checklist:

  • •Verify the implementation exists (check files, endpoints, pages)
  • •Note any items that are incomplete or partially done

3. Run Tests

Execute the appropriate test commands:

bash
make test-unit          # Always run
make test-int           # Always run if integration tests exist
make test-bats          # Run if BATS tests exist for this milestone
make test-playwright    # Run if Playwright tests exist for this milestone

Report pass/fail counts.

4. Verify Use Case Coverage

For each UC listed in the milestone spec:

  • •Check that a corresponding BATS test exists in e2e/cli/*.bats
  • •Check that a corresponding Playwright test exists in e2e/web/*.spec.ts (where applicable)
  • •Flag any UCs without test coverage

5. Check for Pattern Drift

Review new code against CLAUDE.md conventions:

  • •Handlers: Are they thin? No SQL in handlers?
  • •Store: All SQL lives here? Proper error wrapping with sentinel errors?
  • •Tests: Table-driven? Using testify/assert?
  • •Svelte: ID-based state? No object copies in $state? Props interfaces?
  • •API: Endpoints match OpenAPI spec? No hand-written types?

6. Doc/Convention Review

Evaluate whether documentation needs updates based on what was learned:

  • •CLAUDE.md: New patterns discovered? Conventions that should be added?
  • •DESIGN.md: Architecture changes from what was planned?
  • •Milestone specs: Scope changes for upcoming milestones?
  • •OpenAPI spec: Drift from implementation?

If updates are needed, make them.

7. Report

Output a structured report:

code
## Milestone M<N> Review

### Definition of Done
- [x] Item 1 — verified
- [ ] Item 2 — INCOMPLETE: reason

### Test Results
- Unit: X passed, Y failed
- Integration: X passed, Y failed
- BATS: X passed, Y failed
- Playwright: X passed, Y failed

### Use Case Coverage
- Covered: UC_XXXX, UC_XXXX, ...
- Missing CLI test: UC_XXXX
- Missing web test: UC_XXXX

### Pattern Drift
- (any deviations from CLAUDE.md found)

### Doc Updates Made
- (any changes to docs/conventions)

### Recommended Actions
- (items needing attention before next milestone)