bug-investigation

---

name: bug-investigation description: Systematically investigate user-reported bugs and issues, generating detailed reports with root cause analysis, recommended fixes, symptom details, flow diagrams, error message mappings, code analysis, and prioritized fix proposals. allowed-tools: TodoWrite, TodoRead, Read, Write, Edit, Glob, Grep, Bash(git:), Bash(mkdir:), Bash(ls:*), mcp__serena__list_dir, mcp__serena__find_file, mcp__serena__find_symbol, mcp__serena__get_symbols_overview, mcp__serena__search_for_pattern, mcp__serena__find_referencing_symbols, mcp__serena__list_memories, mcp__serena__read_memory, mcp__serena__think_about_collected_information user-invocable: true

Bug Investigation Skill

Systematically investigate user-reported bugs and issues, generating comprehensive analysis reports.

Usage

Copy/bug-investigation [symptom description or Issue number]

• With arguments: Start investigation based on the specified symptoms or Issue number
• Without arguments: Interactively gather symptom information from user

Arguments are available as $ARGUMENTS.

---

Investigation Process

IMPORTANT: Always use Serena MCP for code investigation. It reduces token consumption by 60-80% and efficiently retrieves information through semantic search.

Phase 1: Symptom Collection and Organization

1. Gather symptom details

   • Specific user experience (step-by-step)
   • Error messages (if any)
   • Reproduction conditions (device, OS, browser, steps)
   • Any temporary workarounds

2. Classify symptoms

   Category	Examples
   UI/UX Issues	Hidden buttons, layout problems
   Authentication Issues	Login failures, session expiry
   Functional Failures	Incomplete processing, errors
   Performance Issues	Slow loading, freezing
   Configuration Issues	External service integration problems

Phase 2: Related Code Investigation

1. Check project structure

   CopyUse mcp__serena__list_dir to explore relevant directories
   Use mcp__serena__list_memories to check related memories
   

2. Identify related files

   • Routing/page files
   • Component files
   • API/service files
   • Configuration files

3. Deep dive into code

   CopyUse mcp__serena__get_symbols_overview to understand file structure
   Use mcp__serena__find_symbol to investigate specific functions/classes
   Use mcp__serena__find_referencing_symbols to trace references
   

Phase 3: Root Cause Analysis

1. Visualize processing flow

   • Use ASCII art to illustrate auth flows, data flows, screen transitions
   • Mark problem points with ★

2. Review error handling

   • Create error message to code mapping table
   • Check for unhandled cases

3. Check environment dependencies

   • iOS Safari specific issues (100vh, safe-area, etc.)
   • Browser compatibility
   • External service settings (Firebase, OAuth, etc.)

Phase 4: Report Generation

Output to .tmp/bug-investigation-<date>-<title>.md

---

Output Format

Generate a Markdown report with the following structure in Japanese:

Copy# バグ調査レポート: <タイトル>

作成日: YYYY-MM-DD

## 調査対象

1. [症状1の要約]
2. [症状2の要約]
（報告された症状の数だけ繰り返し）

---

## 1. [症状名]

### 症状の詳細

**ユーザーが経験する状況**:
1. [ステップ1]
2. [ステップ2]
3. [問題が発生]

**ユーザーが見る画面/メッセージ**:
- [具体的な表示内容]

**影響を受ける環境**:
- [デバイス/OS/ブラウザ]

### 処理フローの図解

[開始] ↓ 操作 [処理1] ↓ 条件分岐 ├─ 正常系 → [成功] └─ ★異常系 → [問題発生]

Copy
### 原因分析

**関連ファイル**: `path/to/file.ext`

**問題のあるコード（XX-YY行目）**:
```typescript
// 問題のあるコード

問題点:

なぜこの問題が発生するのか: [技術的な説明]

エラーメッセージと原因の対応表

推奨修正

修正方針: [修正の概要]

Copy// 修正後のコード例

修正の効果:

• [効果1]
• [効果2]

---

2. [次の症状名]

（同様の構造で繰り返し）

---

優先度と対応推奨

---

追加調査が必要な項目

[問題名]について

• 必要な情報: [収集すべき情報]
• 確認事項: [確認すべき項目]

---

関連ファイル一覧

Copy
---

## Investigation Guidelines

### Flow Diagrams

- **Processing flows**: Illustrate auth, data processing, screen transitions
- **Screen structure**: Show UI layout relationships using ASCII art (especially for layout issues)
- **Problem points**: Mark with ★ where issues occur

### Error Message Mapping

- Investigate error handling in code
- Organize mapping between user-visible messages and error codes
- Point out any unhandled cases

### Root Cause Deep Dive

| Aspect | What to Check |
|--------|---------------|
| Code Logic | Missing conditionals, unhandled edge cases |
| Async Processing | Timeouts, race conditions, missing await |
| Environment | iOS Safari, browser compatibility, viewport units |
| External Integration | Firebase, OAuth, API settings |
| CSS/Layout | safe-area, fixed elements, z-index |

### Priority Criteria

| Priority | Criteria |
|----------|----------|
| **High** | Users cannot use the service, security risks |
| **Medium** | Feature works but inconvenient, affects some users |
| **Low** | UX improvements, preventive measures |

### Effort Criteria

| Effort | Criteria |
|--------|----------|
| **Low** | Few lines of changes, config changes only |
| **Medium** | Multiple file changes, testing required |
| **High** | Design changes, large-scale refactoring |

---

## Important Notes

- **Write report in Japanese**: All output content must be in Japanese
- **Specific code references**: Include file paths and line numbers
- **Mark uncertainties**: Use phrases like "可能性がある" (possibly) or "要確認" (needs confirmation) when uncertain
- **Reproduction steps**: Organize reproduction steps whenever possible
- **Request screenshots**: Ask users for screenshots when needed

---

think super hard