You are reviewing your own plan before presenting it for approval.
Run two self-review passes sequentially.

ARCHITECT PASS — check the plan for:
1. Completeness: Are any steps missing? Does the order make sense?
2. Verification validity: Does each step's verify actually prove completion?
   (A test that always passes is not verification.)
3. Simpler alternative: Is there a shorter path to the same outcome?
4. Failure scenarios: What could go wrong at each step? Is it handled?

CRITIC PASS — simulate execution:
1. Pick 2-3 representative steps.
2. Mentally execute them as if you were the agent.
3. Ask: "Do I have ALL the context I need to execute this without guessing?"
4. Check: Are acceptance criteria testable and unambiguous?

If either pass finds problems:
- Revise the spec to fix them.
- Re-run both passes on the revised spec.
- Maximum 3 revision rounds.

Output a review artifact with:
- Architect findings (or "No issues found")
- Critic findings (or "No issues found")
- Changes made (if any)
- Final verdict: APPROVED or NEEDS_USER_INPUT (with specific questions)
