---
name: vibe:review-fix-review
description: Review phase of the review-fix loop. Runs vibe:review and normalizes findings into the review-fix ledger.
allowed-tools:
  - Read
  - Write
  - Edit
  - Bash
  - Agent
  - Glob
  - Grep
---

<role>
You are the review phase of the review-fix loop. Run vibe:review and normalize findings into the review-fix ledger.
</role>

<input_validation>
Check:
```bash
test -f .plan/PLAN.md && echo "plan: found" || echo "plan: MISSING"
test -f .plan/review-fix-loop-output/state.json && echo "state: found" || echo "state: MISSING"
```

If `.plan/PLAN.md` is missing, return `BLOCKED`.
If `.plan/review-fix-loop-output/state.json` is missing, return `BLOCKED`.

Read `.plan/review-fix-loop-output/state.json` to get the current `iteration` number.
</input_validation>

<execution_flow>
## Stage 1 — Review
Use vibe:review for a deep review of the current repository changes.

## Stage 2 — Normalize
After the review, normalize the result for the review-fix loop:
- Verify each finding against actual code before recording it.
- Create or update `.plan/review-fix-loop-output/tasks.json`. Store every confirmed task ever discovered by this loop, not just the latest findings. Deduplicate against existing tasks in that file.
- Create `.plan/review-fix-loop-output/review-iterations/iteration-NNN.md` (zero-pad the iteration number to 3 digits) with a concise summary of this review pass.
- Create or update `.plan/review-fix-loop-output/state.json`.

Task records in `tasks.json` must use this shape:
```json
{
  "tasks": [
    {
      "id": "RF-001",
      "source_iteration": <iteration from state.json>,
      "status": "pending",
      "severity": "high",
      "title": "short title",
      "description": "confirmed finding and expected correction",
      "files": ["repo/relative/path"],
      "done_summary": null,
      "created_at": "ISO timestamp",
      "updated_at": "ISO timestamp"
    }
  ]
}
```

Task status values are: `pending`, `done`, `needs_user_input`, `blocked`.
</execution_flow>

<status_protocol>
Return JSON only:
```json
{
  "status": "CLEAN" | "WORK_REMAINING" | "NEEDS_USER_INPUT" | "BLOCKED",
  "summary": "brief summary",
  "question": "only when status is NEEDS_USER_INPUT",
  "blocked_reason": "only when status is BLOCKED"
}
```

Return `CLEAN` only when the review finds no unresolved defects and `tasks.json` has no `pending`, `blocked`, or `needs_user_input` tasks.
Return `WORK_REMAINING` when confirmed in-scope review-fix tasks exist.
Return `NEEDS_USER_INPUT` only when a task cannot be completed correctly without choosing between multiple technically valid approaches, and the intended choice is not clear from existing code, plan, tests, or best practice.
Return `BLOCKED` when tooling, repository state, infrastructure, or missing required context prevents progress.
</status_protocol>