self-review

name	self-review
description	タスク完了前のセルフレビュー。Gemini CLI + Claude subagent によるマルチ視点検証。

Self Review Skill

タスクを完了（TodoWrite completed）とマークする前にこのセルフレビューを実行。

Trigger Conditions

• TodoWrite でタスクを completed にマークする直前
• ユーザーが /self-review コマンドを明示的に実行した時

Review Procedure

Phase 1: Collect Change Diffs

# 変更ファイル一覧
git diff --name-only HEAD

# 変更内容取得
git diff HEAD

Phase 2: Gemini CLI Review (External Model Perspective)

git diff HEAD -- "*.ts" "*.tsx" | gemini -p "
Please review the following code changes.

## Review Points (Check All)

### 1. Code Quality
- Readability (is it easy to understand?)
- Naming (appropriate variable/function names)
- Structure (proper separation of concerns)

### 2. Bugs & Errors
- TypeScript type safety
- Null/undefined handling
- React hooks rules
- Memory leaks (useEffect cleanup)
- Edge cases

### 3. Design Patterns
- Architecture consistency
- Zustand store patterns
- Repository pattern usage
- Component composition

### 4. React Native Specific
- Performance (FlatList optimization, memo usage)
- Platform differences (iOS/Android)
- Accessibility

## Output Format
- Critical issues: [file:line] [description and fix suggestion]
- Minor issues: [file:line] [description]
- No issues: 'LGTM'
"

Phase 3: Claude Subagent Review

Task ツールで general-purpose エージェントを起動:

subagent_type: general-purpose

Prompt:
---
Please review the following code changes.

## Changed Files
[git diff --name-only HEAD result]

## Change Contents
[git diff HEAD result]

## Review Points

### Common Points
1. Code quality
2. Bugs & errors (TypeScript, hooks, edge cases)
3. Design patterns

### Project-Specific Points
4. Naming (files: camelCase/PascalCase, variables: camelCase, types: PascalCase)
5. Architecture (stores/, repositories/, screens/, components/)
6. React Native specific (performance, accessibility)
7. Zustand patterns (create, get, set usage)

## Output Format
- Critical issues: [file:line] [description and fix suggestion]
- Minor issues: [file:line] [description]
- No issues: 'LGTM'
---

Phase 4: Result Integration

判定	条件	アクション
PASS	両方 LGTM	タスク完了可能
MINOR	軽微な問題のみ	警告表示、タスク完了可能
FAIL	重大な問題あり	修正タスク追加、再レビュー必要

Phase 5: Feedback Loop

FAIL 判定時:

1. 問題箇所を修正
2. Phase 1-4 を再実行
3. PASS になるまで繰り返す

Review Checklist

Code Quality

• 関数は単一責任原則に従う
• 適切なエラーハンドリング
• コメントは最小限で明確
• マジックナンバーは定数定義

TypeScript

• any 型の使用を避ける
• 適切な型定義
• null/undefined の安全な処理

React / React Native

• Hooks rules に従う
• useEffect のクリーンアップ
• 適切なメモ化（useMemo, useCallback, memo）
• FlatList の keyExtractor

Zustand

• Store 構造が適切
• 非同期アクションのエラーハンドリング
• Optimistic update の適切な使用

Output Template

## Self Review Result

### Gemini Review (External Model)
[Output from Gemini]

### Claude Subagent Review (Different Context)
[Output from subagent]

### Comparison Analysis
- Common issues: [Issues identified by both]
- Gemini-only issues: [...]
- Subagent-only issues: [...]

### Judgment: [PASS/MINOR/FAIL]

#### Issues (if applicable)
- [ ] [file:line] [description]

#### Next Action
- [Complete task / Fix required]