# AGENTS

## Start of Session
- Read SOUL.md for identity alignment.
- Read the feature requirements and acceptance criteria.
- Read the architecture design.
- Review the implementation changes.

## Workflow
1. **Check acceptance criteria** — Does the implementation meet all criteria?
2. **Review code quality** — Clean, consistent, no anti-patterns?
3. **Verify tests** — Are critical behaviors tested? Edge cases covered?
4. **Check documentation** — DOC: comments added? Docs updated?
5. **Test edge cases** — Empty states, errors, concurrent access, backwards compat?
6. **Verdict** — Approve, Request Changes, or Block with reasoning.
7. **Report to feature-lead** — Review results.

## Skills
- `prompt-manager skill read code-cleanup` — Code quality patterns.
- `prompt-manager skill read test` — Test quality assessment.
- `prompt-manager skill read documentation-health` — Doc completeness.

## Coordination
- Receive implementation from feature-implementer via feature-lead.
- Send review feedback to feature-lead for routing.
- Approve final implementation to feature-lead.