| name | review-fix |
| description | Fixes blocking issues identified during code review to achieve APPROVED status. Use when addressing review feedback or resolving blocking bugs. Applies minimal targeted fixes, re-validates, and iterates up to 3 times. |
| model | sonnet |
| metadata | {"version":"1.0.0","category":"workflow-automation"} |
Review Fix
Mindset: Get to APPROVED by fixing blocking issues only.
Goal
Fix all blocking issues identified in REVIEW.md so the code can ship.
Instructions
Step 1: Read REVIEW.md
Identify blocking issues:
### Blocking (must fix)
- `src/auth.ts:45` - Missing error handling
- `tests/auth.test.ts` - Test for OAuth failure missing
Step 2: Fix Each Issue
For each blocking issue:
- Read the file at the specified location
- Apply minimal fix
- Run related tests
- Mark as fixed in REVIEW.md
Step 3: Update REVIEW.md
## Issues
### Blocking (must fix)
- ~~`src/auth.ts:45` - Missing error handling~~ FIXED
- ~~`tests/auth.test.ts` - Test for OAuth failure missing~~ FIXED
### Non-Blocking (nice to have)
- `src/utils.ts` - Consider extracting helper (skipped)
Step 4: Run Full Validation
- Run all tests
- Run type check
- Run lint
- Run build
Step 5: Update STATUS.md
Update with fix iteration count and test results.
Fix Rules
DO:
- Fix exactly what's described
- Add tests for bug fixes
- Keep changes minimal
DON'T:
- Refactor unrelated code
- Add new features
- Fix non-blocking suggestions
- Over-engineer the fix
Iteration Limit
Maximum 3 fix iterations:
- Iteration 1: Fix issues
- Iteration 2: Fix remaining issues
- Iteration 3: Last attempt
If still failing after iteration 3:
- Mark as BLOCKED
- Create diagnostic
- Escalate to user
Output Format
- Fixed code files
- Updated
REVIEW.md with struck-through fixed issues
- Updated
STATUS.md with iteration count
Quality Check
Common Issues
- Scope creep: Don't fix non-blocking issues. Focus only on what blocks APPROVED status.
- Over-fixing: Apply minimal changes. Don't refactor surrounding code.
- Iteration overflow: If not fixed by iteration 3, escalate to user instead of continuing.
- New regressions: Always run the full test suite after fixes to catch unintended side effects.