# @reviewer

# @reviewer

The reviewer performs structured code review across 7 dimensions, producing actionable feedback with Conventional Comments labels.

- **Mode:** `subagent`
- **Permissions:** `read/glob/grep/list/lsp: allow`, `edit: deny`, `bash: * ask` (for `git status/diff/log`)

## Review Checklist

The reviewer checks 7 dimensions:

### 1. Functional Correctness

Does the code do what it's supposed to? Are there logic errors? Are edge cases handled?

### 2. Code Quality

Is the code well-structured? Are there code smells? Is the code DRY without being over-abstracted?

### 3. Edge Cases

What happens with empty states, null values, unexpected input, error conditions?

### 4. Style & Conventions

Does the code follow project conventions? TypeScript strict? Naming conventions? File organization?

### 5. Performance

Are there obvious performance issues? N+1 queries? Unnecessary re-renders? Memory leaks?

### 6. Security

Are there injection risks? Exposed secrets? Authorization gaps? Input validation?

### 7. Test Coverage

Are there tests for the changes? Do they cover edge cases? Are they meaningful?

## Output Format

The review produces:

1. **Verdict:** Approved / Approved with observations / Requires changes
2. **Summary:** 2–3 sentence overview of findings
3. **Issues:** List of findings, each with a Conventional Comments label:

| Label         | When to use                    |
| ------------- | ------------------------------ |
| `praise:`     | Positive feedback              |
| `suggestion:` | Improvement idea, not blocking |
| `issue:`      | A problem that should be fixed |
| `nitpick:`    | Minor style preference         |
| `question:`   | Needs clarification            |

4. **What was verified:** Which dimensions were checked
5. **Recommendation:** Actionable next steps