sizukutamago

実装完了後のコードレビューに使用。仕様準拠・コード品質・セキュリティの3観点を並列レビューし、指摘を依存解決して修正する。

セッション完了後の振り返りに使用。ワークフロー遵守の検証、ユーザ指摘の収集と自己改善提案、改善後のリグレッション確認を行う。

ワークフローを開始する。ユースケースを選択し、適切なステップを順に実行する

設計確定後、実装前に使用。設計をタスクに分解し、実装順序と依存関係を明確にする。

大規模プロジェクト（複数機能・フェーズ分割が必要）で使用。全体をフェーズに分割し、マイルストーンとリスクを定義する。各フェーズは planning スキルに渡す。

機能開発・バグ修正の着手前に使用。メインセッションで人間パートナーと対話し、要件を構造化する。コード調査は Explore エージェントに委譲。

要件確定後、実装前に使用。メインセッションで人間パートナーと対話しながら設計を深掘りし、合意した設計を design.md にまとめる。

ハーネスのワークフロー・スキル・エージェント・ルールの使い方を対話的に教える。新メンバーがプロジェクトに参加したときに実行する。

Fetch daily AI news, research papers, and deep dive analysis with self-evolving source registry. Use when user says 'AIニュース', 'ai-news', '論文', 'papers', '最新AI', 'daily digest', '最新の論文', 'arxiv', 'AI最新動向', 'papers today', '今日のAI', '研究動向', 'AI news', '論文まとめ', 'AI業界ニュース', or wants any kind of AI research or industry news digest. Make sure to use this skill proactively when the user mentions AI papers, research trends, or wants to stay updated on AI developments.

Fetch daily business and tech industry news with self-evolving source registry. Use when user says 'ビジネスニュース', 'biz-news', 'テックニュース', 'スタートアップ', 'business news', 'tech news', '資金調達', 'M&A', '業界ニュース', or wants business/tech industry digest. Use proactively when the user mentions startup funding, tech acquisitions, industry trends, or wants to stay updated on business developments.

Claude Code settings catalog — discover what you can customize. Fetches latest settings and env vars from 3 sources (JSON Schema + official docs + official env-vars docs) to show all available options. Use when user says '設定', 'settings', '設定一覧', '設定カタログ', 'what can I configure', '何が設定できる', 'Claude Code設定', '設定項目', '環境変数', 'env vars', or wants to discover available Claude Code configuration options.

Git workflow safety procedures: diagnose why git add fails (3-layer gitignore system), recover from pre-commit hook failures (NEVER amend after hook failure), validate git add -A risks, and manage auto-generated directories. Use when git add fails, pre-commit hook fails, user attempts broad git add, gitignore issues arise, or git commit workflow encounters errors.

Deep-dive into problem clarity before building. Raises resolution on problems using 4 perspectives (depth, breadth, structure, time) through structured dialogue with Why Tree (7+ levels deep). Use when user says '課題を深掘り', '解像度を上げたい', '作る前に整理', 'problem shaping', 'shape the problem', or wants to clarify what to build and why.

End-of-day retrospective: review accomplishments, update task state, and prepare for tomorrow. Use when the user says '振り返り', '今日の成果', '帰る前に', 'daily log', 'wrap up', '1日の終わり', or wants to review what they accomplished today.

Comprehensive deep research that produces structured, multi-section reports with source citations. Only use when user explicitly requests deep/thorough research via '/deep-research' command or says '徹底調査', 'deep research', '調査レポート作成', '包括的なレポート'. Do NOT trigger for simple lookups like '調べて' or 'リサーチして' — use WebSearch or gemini-collab for those.

Blog drafting workflow: create a structured blog post draft from recent work, decisions, and learnings. Use when the user says 'ブログ', 'blog', '記事', 'アウトプット', '書きたい', or wants to turn recent work or learnings into a blog post.

Run the full workflow pipeline. Use when the user wants to "run blueprint workflow", "automate spec to docs", "orchestrate pipeline", "run the full pipeline", "execute blueprint", "start blueprint", "run complete design-to-code pipeline", "spec to implementation", "end to end development", "resume pipeline", "continue from where I left off", "force rerun all stages", "build full-stack app from scratch", "design and implement screens with API", or "create full-stack app with UI". Also use when the user says "パイプラインを実行", "仕様から実装まで", "一気通貫で開発", "設計から実装を自動化", or "フルスタックアプリを作る". Automates /requirements → /spec → /test-from-contract → /implement → /generate-docs with Review Gates (Requirements / Contract / Test / Code / Doc) between each stage. Supports --resume (continue from checkpoint) and --force (rerun all stages).

Define requirements through structured user interviews using Double Diamond pattern and EARS-inspired notation. This skill should be used BEFORE /spec — use /spec when you already know what to build and need I/O boundary contracts, use /requirements when you need to figure out WHAT to build first. Use when the user wants to "define requirements", "gather requirements", "create user stories", "interview for requirements", "identify personas", "define MVP scope", "create acceptance criteria", "structure requirements", "start a new project", "plan a new app", "figure out what to build", or "scope out a product". Also use when the user says "要件定義", "ユーザーストーリー", "要件をまとめる", "ヒアリング", "ペルソナを定義", "MVPを決める", "受け入れ基準", "何を作るか整理", "新しいプロジェクトを始めたい", "アプリのアイデアがある", "何を作るか決めたい", "プロダクトの要件を整理", or "機能を洗い出したい". Make sure to use this skill proactively when the user describes an app idea or project concept without existing contracts or specs.

Analyze accumulated blueprint usage logs and generate improvement PRs for the plugin itself. Use when the user wants to "improve blueprint plugin", "analyze blueprint logs", "self-improve", "check improvement opportunities", "generate improvement PR", "review plugin usage patterns", "blueprint-improve", or "show blueprint stats". Also use when the user says "blueprint改善", "ログ分析", "自己改善", "改善PRを出す", "プラグインを改善する", "使用パターンを分析", or "ログの統計を見せて". This skill reads usage logs from ~/.claude/blueprint-logs/ and identifies recurring patterns, common errors, and user correction frequencies to propose concrete improvements to the blueprint-plugin's core specs, skills, and templates.

Create I/O boundary contracts through brainstorming. Use when the user wants to "create contract", "define spec", "brainstorm API design", "design I/O boundary", "define external integration", "specify file format", "design screen", "create UI spec", "define form contract", "screen contract", "design page layout", or "define frontend spec". Also use when the user asks to start a new feature, create .blueprint directory, or brainstorm a new service. Also use when the user says "仕様を作る", "APIを設計する", "コントラクト作成", "画面を設計する", "I/O境界を定義する", "ブレインストーミング", or "新機能の仕様を決める". Combines interactive brainstorming with structured contract YAML generation for all types: api, external, file, internal, and screen.

Implement code from Contract YAML and RED tests. Use when the user wants to "implement contracts", "generate implementation", "make tests green", "create implementation", "write code from spec", "build feature from contract", "implement feature", "implement screen", "create page components", "implement UI from contract", or "implement Stage 3". Also use when the user says "実装する", "コードを書く", "テストをグリーンにする", "機能を実装する", "画面を実装する", or "コントラクトを実装". Orchestrates Implementers, Integrator, and Refactorer agents to produce working code for all contract types including screen/UI.

This skill should be used when the user asks to "analyze implementation gap", "evaluate feasibility", "assess codebase compatibility", "compare requirements with existing code", or "plan implementation approach". Analyzes gaps between requirements and existing codebase to inform implementation strategy.

Generate design documents from existing code. Use when the user wants to "generate design docs", "create documentation from code", "reverse engineer architecture", "extract design documents", "document existing system", "create docs from implementation", "generate screen design docs", "document UI components", or "create frontend design specs". Also use when the user says "設計書を生成する", "ドキュメントを作る", "コードから仕様書を作成", "画面設計書を生成", or "アーキテクチャドキュメントを書く". Analyzes source code and generates design documentation.

Generate TDD tests from Contract YAML files. Use when the user wants to "generate tests from contracts", "create TDD tests", "test from contract", "generate test stubs", "create Level 1 and Level 2 tests", "write contract tests", "TDD from contracts", "generate UI tests", "create screen tests", "test UI components from contract", or "create frontend test stubs". Also use when the user says "テストを生成する", "コントラクトからテスト作成", "TDDテストを書く", "テストスタブを生成", or "コントラクトをテストする". Reads .blueprint/contracts/ and produces Level 1 (structure validation) and Level 2 (implementation stubs) test files, plus UI tests in tests/ui/ for screen contracts.

Add a new TODO task via natural language chat. AI structures the input into a task file and places it in inbox/. Asks follow-up questions via AskUserQuestion only when context is insufficient for planning. Use when user says "TODO追加", "タスク追加", "todo-add", "やること追加", "〇〇して", "〇〇追加して", or wants to add a task to the auto-todo system.

Approve a planned task for execution. Moves the task file from planned/ to approved/. Use when user says "承認", "approve", "todo-approve", "実行して", "これでOK", or wants to approve a planned TODO task.

Cancel a TODO task. Moves the task file to done/ with cancelled status or deletes it. Use when user says "キャンセル", "取り消し", "cancel", "todo-cancel", "やめる", or wants to cancel a TODO task.

Scheduled task version of todo-execute. Runs daily via Desktop Scheduled Tasks. Executes approved tasks in git worktrees with safety constraints. NOT user-invocable — triggered only by Desktop Scheduled Tasks.

Execute approved TODO tasks. Creates git worktrees for isolation, implements changes according to plans, runs tests, and creates PRs. Use when user says "実行", "execute", "todo-execute", "タスク実行", "これやって", or wants to execute approved tasks immediately.

Scheduled task version of todo-scan. Runs hourly via Desktop Scheduled Tasks. Scans inbox/ for new tasks, analyzes codebases, generates execution plans. NOT user-invocable — triggered only by Desktop Scheduled Tasks.

Manually trigger scan+plan for inbox tasks. Analyzes codebases and generates execution plans, then moves tasks from inbox/ to planned/. Use when user says "スキャン", "scan", "todo-scan", "計画立てて", "タスクを分析", or wants to manually trigger task planning.

Show status summary of all TODO tasks across all directories. Use when user says "タスク一覧", "TODO確認", "todo-status", "状態確認", "何がある", "タスクどうなってる", or wants to see task overview.

Interactive SFRA codebase investigation and exploration skill. Supports direct code investigation (route tracing, override analysis, business logic, data flow, hook investigation) with or without a pre-generated Resolution Map. Triggers on "SFRA explore", "SFRA investigate", "SFRA 探索", "SFRA 調査", "コード調査", "コード探索"

SFRA (Storefront Reference Architecture) code review skill using Swarm pattern. Analyzes controllers, models, ISML, services, jobs for best practices, security, and performance. Triggers on "SFRA review", "SFCC code review", "cartridge review"