setup

config（~/.claude/todo/sources.yaml + sources.local.yaml）に列挙された外部 capture ソース — Slack saved（後で見る）/ 📌 リアクション / self-DM、Google Tasks、 Apple Reminders（remind --dump）、Claude Code transcript の deferral 発話、 カレンダー連携 Notion 議事録 — を sweep し、 TODO 形に正規化（完了条件の起草 + provenance 付与）して機密性 routing どおりの memory store（ai-memory / memory-local）に tags:["todo"] で書き込む毎日ループ skill。 explicit ソース（明示マーク）は自動保存 + 領収書 1 行、inferred ソース（推論抽出）は 候補提案のみ（承認後に保存）。dedup は provenance キー。結果は ~/.claude/todo/ledger.md に追記。設計の真実源は ~/.claude/docs/todo-management.md。 「todo collect」「TODO 集めて」「capture 集めて」「TODO 収集」でトリガー。 注: ソース側の状態は絶対に変更しない（Task の完了化・saved 解除等をしない）。 例外は連携ストアの Apple Reminders のみ（リンクトークン annotate と close 伝播 complete）。 仕事ソースの sink は memory-local 固定 — 個人 ai-memory へ流すのは egress 違反。

全 TODO ストア（各 repo の TODO.md、ai-memory / memory-local の tags:["todo"]、 federated な GitHub issues）を横断列挙し、項目ごとに 2 秒 probe で stale を検出して 済み / 前提消滅 / 継続 / 要判断 に分類する、TODO 管理パイプラインの週次棚卸しループ skill。 probe なしの即答列挙（list モード）も担当。結果は ~/.claude/todo/ledger.md に書き出す。 ストア設計・capture routing は ~/.claude/docs/todo-management.md が真実源。 デフォルトは dry-run（列挙 + probe + 分類提案のみ。forget も TODO.md 編集もしない）。 「todo reconcile」「TODO 棚卸し」「TODO 掃除」で dry-run、 「TODO 見せて」「open な TODO は」で list モード、「LIVE mode」明記で適用モード。 注: forget / TODO.md 編集は機械的証拠か明示承認がある場合のみ。TODO.md 編集は PR 経由（main 直 push 禁止）。issues と TODO.md は列挙するだけで複製しない。

Check the connection and authentication status of every MCP server and re-authenticate any that need it, one at a time. Use when the SessionStart MCP health hook reports "NEEDS-AUTH" / "MCP connectors need attention", when an mcp__* tool call fails with an auth or connection error, or when the user asks to check MCP status, reconnect MCP, "MCP 認証", "check mcp", "mcp 繋がってない", "認証されていない MCP を認証して".

self-heal observability ループの Layer 2-B（issue 解決）。shin1ohno/setup の open な self-heal ラベル issue を 1 件選び、fleet を調査して根本原因を特定し、修正する自律ループ。 allowlist 内の remediation class（既知サービスの再収束／cookbook drift 修正）は 調査→修正→PR→CI green→merge→auto-mitamae 適用→機能検証→issue close まで完全自律。 allowlist 外（新規設計・破壊的変更・auth/secret・原因不明・3 回失敗）は PR/診断を残して self-heal-needs-human を付け停止。merge は CI green 必須、破壊的操作禁止、1 回 1 issue。 owner(shin1ohno) が issue/PR に付けたコメント・更新は GO 承認として拾い、needs-human でも 再着手する（第三者のコメント・更新は無視。bot コメントは <!-- self-heal-bot --> で識別）。 「self-heal resolve」「self-heal issue を直す」「fleet alert を解決」でトリガー。 対は self-heal-create（ES 状態を issue 化する loop）。

Scaffold a new documentation-hub repository for a project: README.md, CLAUDE.md (AI agent instructions), .claude/rules/ discipline files, bundled skills/agents (interview, retro, writing, notion-cli, claude-docs-researcher, session-retrospective, sync-bundled-tools), .gitignore, and an initial git commit. Use when user asks to "set up a docs hub", "create a documentation repository", "新しいプロジェクトのドキュメントハブを作る", "ドキュメント管理リポジトリを scaffold", "init docs project", or starts a fresh repo intended to manage external doc links (Notion, Slack, etc.) rather than source code.

Progressive load testing and performance tuning for Docker services. Identifies hard constraints, finds breaking points, and proposes tuning within safe limits.

平日朝に夜間の開発シグナル（自分の open PR の失敗 CI / レビュー依頼された PR / 自分にアサインされた issue / 夜間 scheduled workflow の失敗）を横断収集し、優先度つきのローカル triage ledger に書き出す Loop Engineering の capstone オーケストレータ skill。CI 失敗には pr-ci-medic を呼ぶ（既定 diagnose）、 well-scoped な修正は worktree で起案し検証サブエージェント（作成/検査分離）に通す。merge は一切しない・出力はローカル。 「朝の triage」「morning triage」「夜間の開発状況まとめて」「daily triage」でトリガー。 注: 自動 merge・自動デプロイはしない。triage と（propose 時の）修正起案までが責務。

Periodically research best practices across software, management, investment, and hobbies. Proposes improvements to this repository.

Research a topic using the memory MCP and web search. Persists findings automatically.

Scaffold release-plz CI + config for a Rust repo on GitHub. Adds the CI workflow, release-plz.toml, enables Actions PR permissions, and guides the CARGO_REGISTRY_TOKEN setup. Use when onboarding a new Rust repo to auto-release via release-plz, or when fixing a repo whose release-plz is failing on a known misconfiguration.

Config-driven web crawler — monitor known sites, run topic-radar searches, or recursively crawl a seed URL; dedups and persists findings to memory (pluggable sink). Run manually as /web-crawl or as a scheduled cloud task via /schedule.

Interactively pick one cookbook from this setup repo and apply it to the local machine via the single-cookbook test harness, with a dependency/ prerequisite preview and a mandatory dry-run before the real apply. Use when the user asks to "apply a cookbook", "apply cookbook", "run a cookbook", "cookbook 適用", "cookbook を選んで適用", "cookbook を適用したい", "pick a cookbook to apply", or names a cookbook to converge locally (e.g. "apply the mise cookbook"). Localhost only — remote/fleet apply (bin/apply-pve-lxcs) is out of scope.

Review the current session and propose improvements to CLAUDE.md, hooks, agents, and skills.

Pre-migration verification for tools targeted at mise / direct-download installs. Runs the 5-check batch (mise registry, GitHub release tag + assets, URL HEAD, .sha256 format, repo language) for each tool before any cookbook line is written. Use BEFORE writing `mise_tool` / `backend "ubi|github|aqua|go|cargo"` / direct curl+install blocks. Trigger phrases - "verify mise backend", "check mise registry", "pre-migration check", "confirm tool specifier", "is X in mise", "does X have darwin asset". MUST be invoked when designing any Homebrew → mise migration or new external-binary-download cookbook entry.

self-heal observability ループの Layer 2-A（issue 作成）。CT111 observer が ES の self-heal-state インデックスに書く open/resolved 状態を、shin1ohno/setup の GitHub issue と 同期する。ES で open かつ未 issue のものを issue 化、ES で消えたものを close + comment、 needs-human は close しない。同期ロジックは決定的な集合演算なので純シェル `self-heal-create.sh` が実装の真実源。この skill はそれを実行して結果を要約するだけ（LLM 判断なし・読み取り専用）。 「self-heal create」「self-heal issue を立てる」「fleet alert を issue 化」「self-heal 同期」でトリガー。 対は self-heal-resolve（issue を解決する loop）。

Document creation and proofreading based on Pyramid Principle + marginal utility. Supports both new creation and editing of existing text.

自分が author の open PR で CI が失敗しているものを巡回し、失敗ログを読んで原因と修正を特定する Loop Engineering の自律 PR/CI 修繕ループ skill。既定は diagnose（原因＋修正案を ledger に記録、コード push なし）、 opt-in で propose（git worktree で修正→ローカル検証→PR ブランチに push、merge はしない）。 merge しない・main に push しない・force-push しない・他人の PR に触れない、が不変の安全境界。 「PR の CI 直して」「failing CI を修正」「pr ci medic」「open PR の赤を見て」でトリガー。 注: merge やリリースは絶対に自動でしない。修繕の提案までが本 skill の責務。

Compare bundled .claude/skills/* and .claude/agents/* against their upstream source repositories and report drift. Primary upstream is github.com/shin1ohno/setup/cookbooks/claude-code/files (for skills/agents originating from that cookbook). External upstreams (e.g. kouzoh/coding-agent-plugins for notion-cli) are handled per the registry below. Use when user asks to "sync skills", "check skill updates", "upstream diff", "cookbook drift", "同梱 skill の差分確認", "上流と同期".

Use this skill when the user wants to compare two implementations and find what's missing — triggers are "compare these two repos", "check parity with the reference SDK", "what features are missing vs X", "audit [my impl] against [reference]", or any port / rewrite / second-implementation context where one codebase should match another. Performs parallel exhaustive inventory of public APIs, methods, events, options, types in both implementations, produces a gap table grouped by priority (high / medium / low), and lets the user pick which gaps to address. Best run in a workspace folder containing both implementations.

Use this skill when the user wants to clarify requirements for a feature, project, decision, or workflow before doing the actual work — typical triggers are "let's spec out X", "interview me about Y", "I want to build Z, ask me what's needed", "before we implement this let's think it through", or any vague request like "make a tool for [thing]" where the scope and constraints are unclear. Runs an 8-dimension AskUserQuestion-driven interview and produces a written SPEC.md the user can review before any implementation begins. Use BEFORE building, not after.

Use this skill when the user wants periodic / quarterly / cross-domain research on best practices — triggers like "what should we adopt", "audit our setup against industry best practices", "research domain X", "AI company practices", or any "look at what's new in [software / management / investment / hobbies] and propose updates to our stuff" framing. Runs domain-specific web research informed by Mem0 user context, prioritizes Anthropic / Claude blog as primary source for AI-tooling updates, surfaces actionable repository / preference / skill changes. Distinct from `research` skill which targets a single topic; this skill spans multiple domains and produces an improvement proposal list.

Use this skill any time the user wants to research, investigate, look up, find out about, or build understanding of a topic — including product comparisons, technical questions, best practices, "what's the state of the art on X", "what do we already know about Y", "compare A and B", or "give me a deep-dive on Z". Searches Cognee (graph + chunks + summaries) and Mem0 (user context) in parallel for prior knowledge, fills gaps via web search, returns a BLUF report with sources, and persists new findings. Distinguishes existing knowledge from new findings, flags single-source claims, and detects contradictions between stores.

Use this skill when the user wants to retrospect on a session, conversation, or recent stretch of work — triggers are "/retro", "let's do a retro", "review this session", "what could we have done better", "any patterns we should codify", or any moment when 3+ commits have been made or the user signals a natural pause point. Analyzes the conversation for repeated corrections, missed AskUserQuestion opportunities, recurring workflows, and effective patterns; proposes concrete updates to user preferences, skills, project memory, or scheduled tasks. Also fires reflexively when the user is "blocked on a manual action" (reading, deciding, restarting) so retrospective work happens during idle time.

Use this skill when the user wants a security review of recent code changes — triggers are "/security-review", "audit this branch for security issues", "OWASP review", "check for vulnerabilities", "is this code safe", or right before merging a PR with non-trivial changes. Diffs the current branch against a base commit, then runs a security-focused review covering OWASP Top 10 (injection, XSS, SSRF, auth flaws, crypto misuse, path traversal, unsafe deserialization), secret leakage, unsafe exec/eval, and untrusted input flowing to privileged sinks. Produces a severity-rated finding list. Requires a git repository in the workspace folder.

Use this skill any time the user wants to write, draft, edit, proofread, restructure, or improve prose — emails, proposals, reports, RFCs, vision documents, blog posts, slack/notion posts, or any text where structure and concision matter. Also triggers when the user asks for "Pyramid Principle" structuring, BLUF rewriting, marginal-utility editing, "make this shorter", "tighten this", "fix the structure", or supplies an existing draft and asks for review. Loads writer + editor personas and optional DVQ/RFC templates; runs a 3-phase Plan → Write → Edit pipeline with up to 3 revision cycles.

Complete a security review of the pending changes on the current branch

Compare implementation against a reference implementation and produce a gap analysis

Check systemd service health on the Linux server and report any failed or degraded units

Interview the user to clarify requirements before building a large feature. Produces a SPEC.md.

Run project tests, lint, and type checks to verify implementation correctness.