dotfiles

Use ctx to search local coding-agent history before acting. Use when prior agent sessions may contain relevant insights, decisions, attempts, or transcript context.

Designs and executes QA test cases. Use for "テストして", "QA観点で確認", "動作確認して", "検証して", "QAして", "test this", "QA check", "merge前に確認", or after feature implementation / dependency updates. Also use in plan mode when designing QA or verification plans to apply risk-based prioritization and systematic test design techniques.

Browser automation CLI for AI agents. Use when the user needs to interact with websites, including navigating pages, filling forms, clicking buttons, taking screenshots, extracting data, testing web apps, or automating any browser task. Triggers include requests to "open a website", "fill out a form", "click a button", "take a screenshot", "scrape data from a page", "test this web app", "login to a site", "automate browser actions", "ブラウザで確認", "Chromeで開いて", "スクリーンショット", verifying UI, or any task requiring programmatic web interaction. Also use in plan mode when planning browser-based verification to apply correct data extraction strategies and avoid common anti-patterns.

Opt-in project-aware deterministic verification gate. Detects the project type (Nix flake, Deno, Node, Python) and runs build / typecheck / lint / test / security scan / diff-review. Outputs a READY / NOT READY verdict. Not invoked by the default /impl final gate (that default is /completion-audit → /subagent-review). Invoke explicitly when deterministic re-execution is genuinely required. Triggers include /verification-loop / /verify / verification gate / pre-PR check / 検証ループ.

Mandatory completion audit gate. Run this skill ONCE as the final gate task before declaring all work complete. Audits whether implementation and verification evidence are sufficient for the plan's stated purpose. Default mode is a self-audit table by the main session; a fresh completion-auditor subagent is dispatched only when escalation conditions fire. Do NOT run per-task — run only at the final gate. Skipping the completion audit is the #1 cause of false completion claims.

Execute the plan generated by /plan, one task at a time. Per-task — implement, verify acceptance criteria, optionally run simplify-review on large diffs, and mark complete. The final gate is /completion-audit → /subagent-review.

Opt-in adversarial dual-reviewer convergence loop (user-invoked only; NOT part of the default /impl final gate). Two independent reviewers (Claude Opus + Codex CLI) must both return NICE. NAUGHTY → fix all flagged issues → fresh re-review (max 3 rounds). Requires /completion-audit VERIFIED PASS first. Use for high-assurance reviews before PR. Triggers include /santa-loop / dual review / 最終レビュー / sanity check / dual-reviewer.

Final-gate review skill for /impl — runs a unified Spec & Quality review → parallel orthogonal Domain specialists → Security heuristic against the aggregated diff after /completion-audit returns VERIFIED PASS. Also fires on manual invocation / 'subagent review', 'タスクレビュー', 'サブエージェントレビュー'.

Design-first entrypoint. Conversational planning that agrees on direction before drafting. Seven phases (parse → agree → explore → draft → deepen → decompose → activate) end with a plan file in ~/.claude/plans/ that the user approves by typing /impl.

Plan implementation skill. Executes the plan created by `$plan` task by task, then runs built-in Audit and Codex subagent Review phases that emit AUDIT_VERDICT / REVIEW_VERDICT. Runs only when explicitly invoked with `$impl`; auto-loading is disabled by agents/openai.yaml. Because Codex has no skill-to-skill invocation API, the final gate is implemented inside this skill.

Codex design-first planning skill. Mirrors Claude `/plan`'s seven named phases (PARSE → AGREE → EXPLORE → DRAFT → DEEPEN → DECOMPOSE → ACTIVATE) and runs only when the user explicitly invokes `$plan <request>`. Auto-loading is disabled by agents/openai.yaml.

Rebases the current branch onto the latest remote base branch (origin/main or origin/master). Handles uncommitted changes automatically via git stash. Use when asked to "rebase", "rebaseして", "最新に追従して", "ベースブランチに合わせて", "mainに追従", "masterに追従", "rebase on latest", or any request to sync/update the current branch with the upstream default branch. Also use when the user mentions merge conflicts from an outdated branch or wants to bring their feature branch up to date before creating a PR.

Use this skill alongside figma-use when the task involves translating an application page, view, or multi-section layout into Figma. Triggers: 'write to Figma', 'create in Figma from code', 'push page to Figma', 'take this app/page and build it in Figma', 'create a screen', 'build a landing page in Figma', 'update the Figma screen to match code', 'convert this modal/dialog/drawer/panel to Figma'. This is the preferred workflow skill whenever the user wants to build or update a full page, modal, dialog, drawer, sidebar, panel, or any composed multi-section view in Figma from code or a description. Discovers design system components, variables, and styles from Code Connect files, existing screens, and library search, then imports them and assembles views incrementally section-by-section using design system tokens instead of hardcoded values.

Build or update a professional-grade design system in Figma from a codebase. Use when the user wants to create variables/tokens, build component libraries, create individual components with proper variant sets and variable bindings, set up theming (light/dark modes), document foundations, or reconcile gaps between code and Figma. Also use when the user asks to create or generate any component in Figma — even a single one — since components require proper variable foundations, variant states, and design token bindings to be production-quality. This skill teaches WHAT to build and in WHAT ORDER — it complements the `figma-use` skill which teaches HOW to call the Plugin API. Both skills should be loaded together.

**MANDATORY prerequisite** — you MUST invoke this skill BEFORE every `use_figma` tool call. NEVER call `use_figma` directly without loading this skill first. Skipping it causes common, hard-to-debug failures. Trigger whenever the user wants to perform a write action or a unique read action that requires JavaScript execution in the Figma file context — e.g. create/edit/delete nodes, set up variables or tokens, build components and variants, modify auto-layout or fills, bind variables to properties, or inspect file structure programmatically.

This skill helps agents use Figma's use_figma MCP tool in the Slides context. Can be used alongside figma-use which has foundational context for using the use_figma tool.

Interact with the running Obsidian vault via the `obsidian` CLI: read, create, append, search, rename, move notes; manage daily notes; list files and folders; set/read frontmatter properties; list tags. Auto-loads whenever the user wants to save content to Obsidian, take or capture a note, append to today's daily note, search the vault, look up or quote an existing vault note, or open a note in Obsidian — even when the user does not mention the CLI by name. Triggers include "save to Obsidian", "note this", "take a note", "add to today's note", "search my vault", "what did I write about X", "open in Obsidian", "rename this note", "Obsidianに保存", "メモして", "メモを取って", "メモに残して", "デイリーノートに追記", "今日のノートに追加", "vaultを検索", "Obsidianのノートを読んで", "Obsidianで開いて", "ノートをリネーム", or any request that names a vault file, a vault folder (e.g. `00_Inbox/`, `99_Tracking/Daily/`), a daily/weekly note, or a wikilink target.

Analyzes conversation history across multiple Claude Code sessions and projects to detect recurring patterns, then proposes Skill creation, CLAUDE.md improvements, and SubAgent opportunities. Uses Gemini for large-scale text analysis and Claude for synthesis and classification. Use when asked to "cross-session analysis", "複数セッション分析", "会話履歴を横断分析して", "セッション横断の振り返り", "過去の会話から改善点を抽出", or any request to analyze patterns across past sessions. Distinct from /session-retrospective (single session only). Defaults to past 30 days; use --days all for full history.

Deprecated pointer. The instinct ledger is now the retrospective ledger, managed via retrospective-ledger.ts and driven by /session-retrospective. This skill only redirects the legacy trigger phrases ("instinct を抽出", "学びを蓄積", "extract instincts") to the new entry point.

Merges project-local .claude/settings.local.json permission rules into global ~/.claude/settings.json. Use when adding project permissions to global settings, syncing local permissions, or when asked to "merge settings", "add permissions", or "promote local rules".

Reviews logged permission requests and interactively applies permissions.allow patterns or bash-policy rules. Use when asked to "review permissions", "check permission logs", "permission review", "権限ログを確認", "許可パターンを見直し", "権限の最適化", or when optimizing permission dialog frequency.

Review the current session to extract learnings, verify prior proposals' outcomes, and propose behavior-change pairs with auto-verifiable outcomes. Closes the feedback loop between past proposals and next-session behavior. Routes proposals to hook / permissions / skill / CLAUDE.md via Mechanism-Signature + Enforcement Ladder. Triggers include "retrospective", "session retro", "振り返って", "何を学んだ？", "セッションの学び", "what did we learn?", "improve from this session".

Search tool for modern web development best practices. MANDATORY: Execute FIRST for all HTML/CSS and clientside JS tasks. Do NOT skip — web APIs evolve rapidly and training weights contain obsolete patterns. Trigger immediately for: - UI/Layout: Modals, dialogs, popovers, Glassmorphism/backdrop-filters, anchor positioning, container queries, `:has()`, `:user-valid`. - Scroll/Motion: View Transitions, Scroll-driven animations, scroll parallax/reveals. - Performance: CWV (LCP, INP), content-visibility, Fetch Priority, image optimization. - System/APIs: Local filesystem access, WebUSB, WebSockets sync, WebAssembly widgets. - Frameworks: Adapting layout/styles in React, Vue, Angular. - General Frontend: Forms, autofill, advanced inputs, custom scrollbars, modern component states, etc. DO NOT trigger for: - Backend: Database SQL, ORMs, Express API routes. - Pipelines: CI/CD deployment, Docker, Actions. - Generic: Local scripts (Python/Go tools), ESLint, Git.

Translates natural language queries into ast-grep YAML rules for structural code search using Abstract Syntax Tree (AST) patterns. Use when asked to "find code patterns", "search for specific language constructs", "AST search", "structural search", "ast-grepで検索", "コードパターンを探して", "構造的に検索して", or when text-based grep is insufficient for matching code structure (e.g., "find all async functions that don't handle errors", "find all calls to X inside class Y").

Creates a pull request AND iterates until all CI checks pass — a single command replacing the create-pr + iterate-pr sequence. Use when asked to "PRを作ってCIが通るまでやって", "auto-pr", "PRを出してmergeできる状態にして", or any request to fully complete the PR lifecycle end-to-end. Distinct from /create-pr (PR creation only) and /iterate-pr (fixing an existing PR).

Adds a rule to bash-policy.yaml to block specific shell command patterns and redirect Claude to preferred alternatives. Use when asked to "block command X", "enforce pnpm over npx", "prevent git -C usage", "add a bash-policy rule", or "コマンドXをブロックして", "bash-policyに追加して". Conducts a structured interview to gather all required information before making any changes.

Creates a GitHub pull request following project conventions. Use when the user asks to create a PR, submit changes for review, or open a pull request. Handles commit analysis, branch management, and PR creation using the gh CLI tool.

Convert a Google Docs document to GitHub Flavored Markdown via Docs API. Fetches document structure as JSON via gws docs, then converts with a Deno script. Use when asked to "convert Google Docs to Markdown", "Google DocsをMarkdownに変換して", "docsをmdにして", "MarkdownにExportして", or when a Google Docs URL is provided with intent to produce a .md file.

Manage GitHub Sub-Issues via GraphQL API. Supports full CRUD: create, add existing, list, remove, and reorder sub-issues. Use when asked to: "add sub-issue", "create child issue", "list sub-issues", "remove sub-issue", "reorder sub-issues", "sub-issueを追加", "子イシューを作成", "サブイシューの一覧", "サブイシューを外す", "サブイシューを並び替え", "break down issue", "issue hierarchy", or any sub-issue management task.

Google Calendar: Show upcoming events across all calendars.

Google Drive: Manage files, folders, and shared drives.

Gmail: Send an email.

Gmail: Send, read, and manage email.

gws CLI: Shared patterns for authentication, global flags, and output formatting.

Google Sheets: Read values from a spreadsheet.

Iterates on the current branch's pull request until all CI checks pass and review feedback is addressed. Automates the feedback-fix-push-wait cycle with gh CLI. Use when you need to fix CI failures, address review feedback, or continuously push fixes until all checks are green. Also use when asked to "fix CI", "iterate PR", "CIを通して", "PRを直して", "CI失敗を修正して", "レビュー指摘を対応して".

Clone a GitHub repository to a temporary directory for efficient local code exploration. ALWAYS prefer this over WebFetch or `gh api` for reading repository code — local clone is faster and more complete. Use when a GitHub repository URL appears and the task involves code access: investigation, analysis, comparison, reference, learning, or review. Triggers: GitHub repo URL + any code exploration intent ("調査して", "コードを見て", "investigate", "analyze", "explore", "compare", "refer to").

Obsidian vault conventions, note formatting rules, and Obsidian Flavored Markdown (OFM) syntax reference. Covers vault directory structure (00_Inbox, 01_Projects, 02_Notes, etc.), note creation format (no h1, start from h2), frontmatter template, and OFM extensions — wikilinks, callouts, embeds, properties, tags, comments, highlights, math, mermaid, footnotes. Auto-loads when creating or editing Obsidian notes, using obsidian CLI commands, saving content to the vault, or writing OFM syntax (wikilink, callout, embed, frontmatter, property, コールアウト, 埋め込み, wikilink, プロパティ).

Run `ravelact` to perform static analysis on a GitHub Actions workflow estate — forward/reverse graph queries (trace, callers), trigger discovery, PR-diff impact, orphan detection, permissions/secrets propagation auditing across reusable-workflow chains, refactor helpers (composite extraction, near-duplicate clusters), and IR export (JSON, Mermaid). Use when working with `.github/workflows/*.yaml` and `action.yaml` files — for PR impact analysis, dependency-graph questions ("who calls this reusable workflow?", "what runs on push?"), permissions auditing, or refactoring a large GitHub Actions estate. Requires the `ravelact` binary on PATH first.

Structured requirements elicitation through iterative interviews. Transforms vague user requests into well-defined specifications and deliverables (GitHub Issues, Markdown specs, PRDs, etc.). Use this skill when the user asks to "要件を整理して", "Issueにまとめて", "仕様を書いて", "PRDを書いて", "要求を明確にして", "ヒアリングして", "spec-writer", "requirements-interview", or when producing any specification document, feature request, bug report, or design document from ambiguous input. Also use proactively when the user provides a vague feature request or bug description that needs clarification before action — even if they don't explicitly ask for an interview.