tumf

アプリ組み込み AI チャットフロントエンドの設計・実装ガイド。 前半は言語非依存の抽象設計パターン（ストリーミングファースト、チャット状態マシン、 メッセージモデル、楽観的UI、エラーリカバリ、UX原則、マルチセッション、サーバー設計原則）、 後半は Vercel AI SDK (useChat) による TypeScript 具体実装。 フレームワーク非依存（Next.js, React SPA, Vue, Svelte, SvelteKit, Nuxt 対応）。 Use when チャット機能を追加したい、AIチャットUIを作りたい、useChat を使いたい、 ストリーミングチャットを実装したい、チャットボットUIを組み込みたい、 AI SDK でチャットを作る、またはチャットUIの設計パターンを整理・レビューしたい時。 Python/Go/Rust 等の非 TypeScript 環境でも設計パターン部分は適用可能。 既存のチャットUIのレビュー・改善にも使える。 対象外: チャット以外の AI SDK 用途（テキスト生成のみ、画像生成、エンベディング等）。

Use this skill whenever the user asks to create, structure, review, or refine a business plan, new business proposal, growth strategy, GTM outline, or venture hypothesis. Build a rational, persuasive business plan with reader-specific framing and bottom-up numerical planning. Distinguish researched facts from benchmarks and estimates, use frameworks only as supporting tools, and forecast outcomes from causal drivers instead of reverse-engineering a desired result. Use it even when the user only says "事業計画", "戦略を考えたい", "新規事業を整理したい", or asks for AI agent / SaaS business planning without naming any framework.

デジタル庁のダッシュボードデザイン実践ガイドブックに基づく、見やすいダッシュボード設計・作成のガイド。グラフ選び、レイアウト、カラーパレット、アクセシビリティ、チェックリストを網羅。Use when ダッシュボードを新規作成したい、既存ダッシュボードを改善したい、可視化設計をレビューしたい、または軽量HTML・Grafana・SQLiteベースのダッシュボード設計方針を決めたい時。

Use this skill whenever the user asks for brainstorming, ideation, concept expansion, service improvement ideas, strategy options, workshop structure, or says thinking is stuck and needs more perspectives. Select and apply one or more brainstorming frameworks based on the user's goal, solo vs group setting, divergence vs convergence need, and business vs creative context. Use it even if the user does not name a framework.

Design principles for building CLI tools that LLM/AI agents can safely and reliably operate. Provides 7 core principles (Machine-readable, Non-interactive, Idempotent, Safe-by-default, Observable, Context-efficient, Introspectable) with scorecard, templates, and anti-patterns. Use when: (1) designing new CLI tools for agent use, (2) improving existing CLIs for agent compatibility, (3) reviewing CLI design for agent-friendliness, or (4) need guidance on JSON output formats, exit codes, authentication flows, or self-describing CLI patterns.

Fast, script-friendly CLI for Google services via `gog`. Use this skill whenever the user needs to work with Gmail, Google Calendar, Drive, Docs, Sheets, Slides, Forms, Apps Script, Contacts, Tasks, People, Chat, Classroom, Groups, Admin, or Keep from the terminal, especially for Google Workspace automation, account setup, OAuth/client configuration, service accounts, domain-wide delegation, or JSON/scriptable workflows. Prefer this skill even when the user does not mention `gog` explicitly but is clearly trying to automate or inspect Google services from a CLI or agent workflow.

Audit and reorganize an existing test suite to remove outdated tests, reduce duplication, and isolate external dependencies from unit tests. Use this skill whenever the user wants to clean up tests, find obsolete or redundant tests, detect unit tests that directly hit APIs/URLs/timers/commands/databases/filesystems/env state, improve mock boundaries, reduce flakiness, or clarify the boundary between unit and integration tests. Use it even if the user only says things like "整理したい", "モック化したい", "テストが遅い", or "古いテストを掃除したい".

Create and maintain `.wt/setup` bootstrap scripts for git worktree environments. The `.wt/setup` script is a repository-local convention executed automatically by supporting tools (e.g. wt) when creating new git worktrees. Use when: (1) a project needs a `.wt/setup` script, (2) setting up worktree-aware bootstrap for a new repo, (3) reviewing or fixing an existing `.wt/setup`, or (4) the user mentions "wt setup", "worktree setup", ".wt/setup", or "bootstrap worktree".

Archive deployed OpenSpec changes and update canonical specs. Provides archive-specific guidance for Conflux orchestration. CRITICAL - This skill CANNOT ask questions or request user input.

Run the standard Conflux development flow for an already-defined and committed OpenSpec change. Use when users want to execute `cflx run`, start Conflux orchestration, or follow the standard proposal-then-run workflow on a clean base branch.

Portable Conflux acceptance operation skill. Defines the JSON-primary verdict interface and autonomous acceptance review guidance for any agent runtime. CRITICAL - This skill CANNOT ask questions or request user input.

Portable Conflux acceptance review with an additional SPECA-style property/proof-attempt lens. Drop-in compatible with cflx-accept and cannot ask questions or request user input.

Implement an approved OpenSpec change autonomously with truthful task tracking. Provides apply-specific guidance for Conflux orchestration. CRITICAL - This skill CANNOT ask questions or request user input.

Legacy compatibility router for Conflux workflow operations. Routes apply/rejecting/cleanup-review/accept/archive to self-contained operation guidance. New orchestrator prompts should use dedicated operation-specific skills (cflx-apply, cflx-accept, etc.) instead. CRITICAL - This skill CANNOT ask questions or request user input.

Create structured Conflux change proposals through interactive conversation with users. Use when users request "create a proposal", "draft a change", "propose a feature", or similar proposal creation tasks. This skill asks clarifying questions and guides users through the proposal process.

Review apply-generated rejection proposals (REJECTED.md) before any terminal reject decision. Returns REJECTION_REVIEW CONFIRM, RESUME, or BLOCK. CRITICAL - This skill CANNOT ask questions or request user input.

Run the standard Conflux development flow for an already-defined and committed OpenSpec change. Use when users want to execute `cflx run`, start Conflux orchestration, or follow the standard proposal-then-run workflow on a clean base branch.

Create structured Conflux change proposals through interactive conversation with users. Use when users request "create a proposal", "draft a change", "propose a feature", or similar proposal creation tasks. This skill asks clarifying questions and guides users through the proposal process.

Execute Conflux workflow operations autonomously without user interaction. Provides three operations - apply (implement approved changes), accept (verify implementation), and archive (finalize deployed changes). Called by Conflux orchestration system. CRITICAL - This skill CANNOT ask questions or request user input.

Interact with X/Twitter via the xcom-rs CLI (Rust). Use for posting tweets, replies, threads, searching, reading timelines/mentions, liking, retweeting, bookmarks, media upload, and user lookups. Use this skill whenever the user wants to do anything on X/Twitter — posting, reading, searching, monitoring mentions, managing bookmarks, or looking up users and their tweets.

Use `agent-exec` whenever shell work may run longer than a safe inline Bash call, may produce more output than you want in context, or should remain observable as a managed job. Start with plain `agent-exec run -- <command>` and rely on its default behavior; do not add custom wait or timeout flags unless there is a concrete reason. This skill is especially relevant when the user wants to run builds, tests, servers, data jobs, or any command whose duration or output size is uncertain.

Slack Web API automation via the slack-rs CLI (Rust). Use when you need to authenticate to Slack via OAuth (PKCE), manage multiple workspace profiles, call arbitrary Slack Web API methods (e.g. chat.postMessage, conversations.list, users.info), and run safe scripted Slack operations from the terminal. Includes tunnel-assisted remote login (see auth login --help), encrypted profile export/import, and a write-safety guard via SLACKCLI_ALLOW_WRITE. Credentials are stored in file-based storage under ~/.config/slack-rs/.

Guidance for agents that use review-gauntlet as a tool to perform coverage-tracked code reviews. Use when asked to run, continue, inspect, verify, or finalize a review-gauntlet review session; triage findings; check unreviewed/stale coverage; or operate review-gauntlet through an external review tool adapter.

Guide for integrating the agent-skills-rs Rust library and optional CLI workflow into a project. Use when adding embedded skill discovery/installation, lock-file updates, agent-specific install paths, or command introspection support in Rust codebases.

Learn reusable per-site web scraping adapters and turn them into lightweight commands for repeated extraction. Use when Claude needs to scrape the same domain more than once, reduce token and HTML parsing cost, validate DOM drift, retrain site-specific rules, or export a learned website as a CLI-style tool instead of re-exploring raw pages every time.

Comprehensive Clawdbot configuration and skills system management. Use when working with Clawdbot setup, configuration files, creating/modifying AgentSkills-compatible skills with YAML frontmatter, troubleshooting agent behavior, managing channels, workspace, sandbox, or multi-agent routing.