home / skills / shunsukehayashi / miyabi-claude-plugins / issue-analysis

issue-analysis skill

safe

/miyabi-full/skills/issue-analysis

This skill analyzes GitHub issues and infers Miyabi's 57-label system to aid triage and labeling efficiency.

npx playbooks add skill shunsukehayashi/miyabi-claude-plugins --skill issue-analysis

Review the files below or copy the command above to add this skill to your agents.

Files (2)

SKILL.md

7.8 KB

---
name: Issue Analysis with Label Inference
description: Analyze GitHub Issues and automatically infer appropriate labels from Miyabi's 57-label system across 11 categories. Use when creating or triaging Issues, or when label inference is needed.
allowed-tools: Read, Grep, Glob, WebFetch
---

# 🏷️ Issue Analysis with Label Inference

**Version**: 2.0.0
**Last Updated**: 2025-11-22
**Priority**: ⭐⭐⭐⭐ (P1 Level)
**Purpose**: AI駆動のIssue分析と57ラベルシステムからの自動推論

---

## 📋 概要

Miyabiの57ラベルシステム（11カテゴリ）に基づいたAI駆動のIssue分析と
自動ラベル推論を実行します。

---

## 🎯 P0: 呼び出しトリガー

| トリガー | 例 |
|---------|-----|
| ラベル質問 | "what labels should I use?" |
| Issue分析 | "analyze this Issue", "triage issue #270" |
| ラベル推論 | "infer labels for this Issue" |
| Issue作成後 | Issue作成後の自動処理 |
| バックログ管理 | "prioritize backlog" |

---

## 🔧 P1: ラベルカテゴリ一覧

### 11カテゴリ・57ラベル

| # | カテゴリ | 数 | 必須/任意 | 用途 |
|---|---------|-----|---------|------|
| 1 | **STATE** | 8 | 自動 | ライフサイクル管理 |
| 2 | **AGENT** | 6 | 推奨 | Agent割り当て |
| 3 | **PRIORITY** | 4 | 必須 | 優先度管理 |
| 4 | **TYPE** | 7 | 必須 | Issue分類 |
| 5 | **SEVERITY** | 4 | 条件 | 重大度（バグのみ） |
| 6 | **PHASE** | 5 | 推奨 | プロジェクトフェーズ |
| 7 | **SPECIAL** | 7 | 条件 | 特殊フラグ |
| 8 | **TRIGGER** | 4 | 自動 | 自動化トリガー |
| 9 | **QUALITY** | 4 | 自動 | 品質スコア |
| 10 | **COMMUNITY** | 4 | 推奨 | コミュニティ |
| 11 | **HIERARCHY** | 4 | 自動 | Issue階層 |

---

## 🚀 P2: ラベル推論ルール

### TYPE推論（必須・1個）

| キーワード | ラベル | 例 |
|-----------|--------|-----|
| "add", "implement", "create", "new" | `✨ type:feature` | 新機能追加 |
| "fix", "crash", "error", "broken" | `🐛 type:bug` | バグ修正 |
| "docs", "README", "guide" | `📚 type:docs` | ドキュメント |
| "refactor", "cleanup", "reorganize" | `🔧 type:refactor` | リファクタ |
| "test", "coverage", "e2e" | `🧪 type:test` | テスト |
| "architecture", "system design" | `🏗️ type:architecture` | アーキテクチャ |
| "deploy", "CI/CD", "docker" | `🚀 type:deployment` | デプロイ |

### PRIORITY推論（必須・1個）

| 条件 | ラベル | SLA |
|------|--------|-----|
| セキュリティ、本番障害、データ損失 | `🔥 priority:P0-Critical` | 24時間 |
| 主要機能、重大バグ、性能劣化 | `⚠️ priority:P1-High` | 3日 |
| 通常機能、標準バグ | `📊 priority:P2-Medium` | 1週間 |
| 軽微改善、typo | `📝 priority:P3-Low` | なし |

### SEVERITY推論（条件付き・バグのみ）

| 条件 | ラベル | エスカレーション |
|------|--------|----------------|
| 本番停止、セキュリティ侵害 | `🚨 severity:Sev.1-Critical` | Guardian + CISO |
| 主要機能障害 | `⚠️ severity:Sev.2-High` | TechLead |
| 部分的機能問題 | `📊 severity:Sev.3-Medium` | Agent自動 |
| 軽微UI問題 | `📝 severity:Sev.4-Low` | Agent自動 |

### SPECIAL推論（条件付き）

| キーワード | ラベル | アクション |
|-----------|--------|-----------|
| "XSS", "SQL injection", "CVE" | `🔐 security` | CISO通知、Issue非公開 |
| "Claude API", "high cost" | `💰 cost-watch` | 予算監視 |
| "depends on #", "blocked by" | `🔄 dependencies` | 依存解決まで待機 |
| "research", "investigate" | `🎓 learning` | SLA延長 |
| "experiment", "POC" | `🔬 experiment` | 失敗許容 |

### COMMUNITY推論（推奨）

| 条件 | ラベル |
|------|--------|
| 2時間以内、依存なし、明確な要件 | `👋 good-first-issue` |
| 外部専門知識が必要 | `🙏 help-wanted` |

### HIERARCHY推論（自動）

| 条件 | ラベル |
|------|--------|
| 親Issueなし | `🌳 hierarchy:root` |
| 子Issueあり | `📂 hierarchy:parent` |
| 親Issueあり | `📄 hierarchy:child` |
| 子Issueなし（末端） | `🍃 hierarchy:leaf` |

---

## ⚡ P3: 分析ワークフロー

### Step 1: Issue内容読み取り

```
- タイトル
- 本文（説明）
- コメント（あれば）
- 関連Issue（リンク）
```

### Step 2: キーワード抽出

```
- 技術キーワード: "security", "performance", "API"
- アクションキーワード: "add", "fix", "refactor"
- 緊急度キーワード: "urgent", "critical", "nice-to-have"
```

### Step 3: 推論ルール適用

```
1. TYPE決定（必須、1個）
2. PRIORITY決定（必須、1個）
3. SEVERITY決定（バグ/インシデント時）
4. SPECIAL条件チェック
5. COMMUNITY適合性評価
6. HIERARCHY位置決定
```

### Step 4: ラベルセット生成

```json
{
  "required": ["type:feature", "priority:P1-High"],
  "recommended": ["agent:codegen", "phase:planning"],
  "optional": ["security", "cost-watch"],
  "automatic": ["state:pending", "hierarchy:leaf"]
}
```

---

## 📊 出力フォーマット

### JSON形式

```json
{
  "issue_number": 270,
  "title": "XSS vulnerability in comment form",
  "analysis": {
    "type": "bug",
    "priority": "P0-Critical",
    "severity": "Sev.1-Critical",
    "estimated_time": "4-6 hours",
    "complexity": "medium",
    "agent_recommendation": "codegen + review",
    "escalation": "Guardian + CISO"
  },
  "labels": {
    "required": ["type:bug", "priority:P0-Critical"],
    "recommended": ["severity:Sev.1-Critical", "security"],
    "automatic": ["state:pending", "hierarchy:leaf"]
  },
  "rationale": {
    "type": "Keywords: 'vulnerability', 'XSS'",
    "priority": "Security issue with global impact",
    "severity": "Security vulnerability"
  }
}
```

### テキスト形式

```
🏷️ Issue Analysis Results

📋 Issue #270: XSS vulnerability in comment form

📊 Analysis:
- Type: bug (Keywords: 'vulnerability', 'XSS')
- Priority: P0-Critical (Security, global impact)
- Severity: Sev.1-Critical (Security vulnerability)
- Estimated: 4-6 hours
- Agent: codegen + review

🏷️ Labels:
✅ Required: type:bug, priority:P0-Critical
📎 Recommended: severity:Sev.1-Critical, security
🔄 Automatic: state:pending, hierarchy:leaf

⚠️ Escalation: Guardian + CISO (immediately)
```

---

## 🛡️ エラーハンドリング

### キーワード検出失敗

```
# デフォルトフォールバック
TYPE: type:feature (不明な場合)
PRIORITY: priority:P2-Medium (不明な場合)
```

### 複数TYPE該当

```
# 優先順位で決定
bug > security > feature > docs > test > chore
```

### コンフリクトするラベル

```
# 相互排他ラベルのチェック
state:pending ⊕ state:in-progress
priority:P0 ⊕ priority:P1
```

---

## ✅ 成功基準

| チェック項目 | 基準 |
|-------------|------|
| TYPE推論 | 1個選択 |
| PRIORITY推論 | 1個選択 |
| SEVERITY推論 | バグ時のみ |
| SPECIAL検出 | 該当時のみ |
| 根拠提示 | 各ラベルに理由 |
| エスカレーション | 必要時に推奨 |

---

## 🔗 関連ドキュメント

| ドキュメント | 用途 |
|-------------|------|
| `docs/LABEL_SYSTEM_GUIDE.md` | ラベルシステム詳細 |
| `.github/labels.yml` | ラベル定義 |
| `agents/specs/coding/issue-agent.md` | IssueAgent仕様 |
| `docs/ENTITY_RELATION_MODEL.md` | エンティティ関係 |

---

## 📝 関連Skills

- **Agent Execution**: ラベル割り当て後のAgent実行
- **Git Workflow**: Issue解決後のコミット
- **Rust Development**: 実装品質チェック

Overview

This skill analyzes GitHub Issues and infers appropriate labels from Miyabi’s 57-label system across 11 categories. It helps when creating, triaging, or prioritizing Issues by producing a consistent label set and a short rationale. Results include required, recommended, optional, and automatic labels plus escalation recommendations when applicable.

How this skill works

The skill reads the issue title, body, comments, and linked issues to extract action, technical, and urgency keywords. It applies deterministic rules to select one TYPE and one PRIORITY, conditionally assigns SEVERITY for bugs, checks SPECIAL and COMMUNITY signals, and determines HIERARCHY and STATE labels automatically. Output is a structured label set with rationale, estimated effort, agent recommendation, and escalation steps.

When to use it

When creating a new Issue to auto-suggest labels
During triage to standardize labels across the backlog
For backlog prioritization and SLA assignment
When security or cost signals may require escalation
When you need an audit-friendly rationale for labels

Best practices

Provide a clear title and concise description to improve keyword detection
Reference related issues or PRs to enable hierarchy detection
Flag security or data-sensitivity details explicitly (e.g., XSS, CVE) for proper escalation
Review inferred TYPE and PRIORITY if the issue mixes multiple intents
Use recommended agent suggestions to speed up resolution and handoffs

Example use cases

Automatic labeling of a newly opened bug with severity and escalation steps
Triaging a feature request into type:feature, priority:P2-Medium, and phase:planning
Detecting security keywords to add security and priority:P0-Critical with CISO escalation
Identifying good-first-issue candidates for community contributors
Assigning hierarchy labels when issues reference parent or child items

FAQ

What happens if keywords are ambiguous?

The skill falls back to type:feature and priority:P2-Medium by default and reports the ambiguity in the rationale.

How many TYPE labels will it assign?

Exactly one TYPE is chosen per rules; if multiple match, a priority order resolves conflicts (bug > security > feature > docs > test > chore).

Does it automatically notify security or CISO?

It recommends escalation and marks security-sensitive issues; actual notifications depend on your repository’s configured workflows and integrations.

Can I customize the label rules?

Rules are deterministic and intended for consistency; customization is supported by updating the label mapping and inference triggers in your configuration.