Art & Design

어떤 도메인의 프로젝트든 하네스 구조로 설계·구축하는 메타 스킬. Layer 1(메타): Planner→Generator→Evaluator 순환으로 워크스페이스를 설계한다. Layer 2(운영): 완성된 구조 안에서 도메인 에이전트들이 실제 업무를 수행한다. 두 레이어는 완전히 분리된다. Evaluator 90점+ PASS 후 cartesian-review 스킬을 자동 호출해 데카르트 5렌즈(열거·명증·종합·분석·방법적 회의)로 추가 품질 검증을 수행한다. cartesian-review 스킬이 설치되어 있어야 이 단계가 활성화된다. "하네스 만들어줘", "워크스페이스 설계", "에이전트 시스템 구축", "CLAUDE.md 만들어줘", "프로젝트 구조 잡아줘", "Council 구조 만들어줘", "반복 가능한 시스템", "자동화 워크플로우", "멀티 에이전트", "독립 분석", "다각도 분석", "build a harness", "design workspace", "agent system" 표현 시 반드시 사용.

UI 以外のユーザー接点(API・設定ファイル/設定項目・拡張機能/プラグイン)を、UX の 6 原則に沿って設計・レビューする。API 設計、REST/SDK/クライアントライブラリの使いやすさ、config.yml や環境変数や設定項目の追加、拡張ポイント/プラグイン機構の設計、「設定を足すべきか」「デフォルト値をどうするか」「APIのバージョニング」「開発者体験(DX)」「オンボーディング」「驚き最小の原則」などの話題が出たら、明示的に UX と言われなくても必ず使う。システム設計の初期段階で UX を先に決めたいときにも使う。

코딩 에이전트가 코드를 안정적으로 생성·테스트·배포할 수 있는 하네스 구조를 설계하는 메타 스킬. Layer 1(메타): Planner→Generator→Evaluator 순환으로 코딩 하네스를 설계·구축·검증한다. Layer 2(운영): 완성된 하네스 안에서 코딩 에이전트가 실제 개발을 수행한다. "코딩 하네스 만들어줘", "개발 워크스페이스 설계", "에이전트 코딩 환경 구축", "CLAUDE.md 코딩 프로젝트용", "CI 파이프라인 설계", "코드 리뷰 루프 구축", "자율 코딩 환경", "에이전트 개발 환경", "coding harness", "agent-first development", "build harness for coding", "dev workspace setup", "autonomous coding environment" 표현 시 반드시 사용.

데카르트 방법서설의 4규칙(열거·명증·종합·분석)에 방법적 회의를 더한 5렌즈로 어떤 문서·설계·계획이든 체계적으로 검토해 실행 가능한 결론을 도출하는 분석 스킬. 다음 표현이 나오면 반드시 이 스킬을 사용한다: - "데카르트 분석", "데카르트 관점으로", "방법적으로 검토" - "네 가지 관점으로 검토", "4가지 관점 분석" - "더하면 좋은 것 / 빼면 좋은 것 / 빼면 안 좋은 것 / 더하면 안 좋은 것" - "빠진 것 찾아줘", "군더더기 찾아줘", "핵심이 뭔지 봐줘" - "다른 관점에서 분석해줘", "외부 시각으로 봐줘" - "이 문서 체계적으로 검토해줘", "이 설계 분석해줘" - "/cartesian-review" - 하네스 Evaluator 단계에서 자동 호출 (harness-designer-workspace 통합) 호출 전에 사용자 의도를 3가지 관점(①검토 대상: 코드·문서·설계·의사결정 / ②검증 깊이: 경량·표준·엄격 / ③목적: 빠진 것 찾기·개선 제안·최종 승인)으로 간단히 확인한 뒤 진행한다. 사용자가 이미 명시한 경우 확인을 생략할 수 있다.

Makes Arabic output correct the first time. Covers right-to-left layout, connected letter shaping, mixed Arabic/English/number ordering, Arabic fonts and typography, numerals, dates, currency, and natural Arabic wording. Use it whenever the output will contain Arabic text or is meant for Arabic speakers. That includes websites and apps (HTML, CSS, Tailwind, React, Next.js, Vue, React Native, Flutter), Word/PowerPoint/PDF files, charts and images (matplotlib, Pillow, SVG, canvas), emails, SMS/WhatsApp/social posts, markdown, and plain text. Use it when the request is written in Arabic (e.g. "صمم صفحة", "اكتب رسالة", "تقرير", "بالعربي"). Use it when an English request mentions Arabic, RTL, right-to-left, bilingual English/Arabic, i18n for Arabic, or users in Saudi Arabia, the UAE, Egypt, the Gulf, the Levant, North Africa or the MENA region. Also use it to fix broken Arabic: reversed or disconnected letters, boxes instead of glyphs, wrong alignment, or punctuation and numbers on the wrong side.

Use this skill when building AI agents, designing multi-agent systems, integrating LLMs into applications, building RAG pipelines, working with vector databases, implementing MCP servers, or designing agent communication and orchestration patterns. Trigger on keywords like: agent, LLM, RAG, embeddings, vector DB, orchestration, tool use, memory, intent clarification, MCP, LangChain, LangGraph, CrewAI, AutoGen, Microsoft Agent Framework.

Use this skill when designing how an AI agent should communicate with developers — intent clarification, progress reporting, ambiguity handling, confirmation before irreversible actions, error communication, self-correction, task tracking, and plan mode. Also use when building agent UX or defining agent behavior rules. Trigger on keywords: agent communication, intent clarification, plan mode, agent behavior, task tracking, confirmation, agent UX, agentic workflow, agent protocol.

Extracts the visual system from a reference site or pasted HTML (colour roles, type scale, spacing scale, radii, shadows, button/card/eyebrow anatomy, tone) and maps it onto the current project's existing tokens as a short token table. Use when the user shares a reference ("make it look like <site>", "in this style", a URL or HTML paste) before a redesign.

Interview the user about something they are ABOUT TO BUILD, before any of it is built — a feature, a system, an automation, a workflow, a document structure. Use when a plan exists only in their head or as a rough idea and the design decisions are still open: \"grill me\", \"interview me about this\", \"help me think this through before I build it\", \"what am I not deciding\", \"ask me what you need to know\". Works the design as a dependency tree, asking one answerable question per round (the one that unblocks the most decisions first) with a recommended answer attached, and never asking anything it could look up itself. BOUNDARY — this is front-loaded, for decisions not yet made. Reviewing a plan that already exists and stress-testing it for holes is the plc skill. Interrogating information someone else handed you is critical-thinking-walkthrough. If the thing is already written down and the question is \"is this any good\", it is not this skill.

Use this skill when designing or reviewing APIs — REST, GraphQL, tRPC, or gRPC. Trigger on keywords: API design, REST, GraphQL, tRPC, gRPC, endpoint, schema, route, OpenAPI, Swagger, API versioning, pagination, API contract, HTTP methods.

Find and run a project's test, lint, type-check, build and dev-server commands, read the results correctly, and seed a throwaway database when UI work needs realistic data. Never seeds the project's real or default database. Use before 'run the tests', 'lint this', 'does it build', 'start the app so I can see it', 'seed some demo data', and before you claim that a change works. Not for deployment or CI pipeline design; use the project's own deploy tooling for those.

Review Go (Golang) code changes like a careful senior engineer, following Google's code review practices and Google's Go Style Guide best practices (design, functionality, complexity, tests, naming, errors, comments, style, docs) plus Go-specific checks for errors, concurrency, context and resource handling. Use this whenever the user asks to review Go code, a diff, a PR/MR, a branch or a .go file, or asks things like "check my changes", "is this Go code okay", "LGTM?", even if they don't say "code review".

Applies an agreed visual direction to a page or component by reusing existing components first, then verifies every width and dark mode with the headless-Chrome audit plus a before/after pixel diff, and reviews against Refactoring UI, Nielsen and WCAG 2.2. Use when asked to redesign, restyle, polish or "make it look like the reference", after design-ref or when the direction is already clear.

>- Extract a comprehensive design system (DESIGN.md) directly from frontend source code — React, Vue, Svelte, Angular, plain HTML/CSS, or any web framework. Analyzes component files, stylesheets, Tailwind configs, theme definitions, and design tokens to produce a rich, Stitch-compatible design system document. Use this skill whenever the user wants to reverse-engineer a design system from an existing codebase, audit the visual language of a project, extract design tokens from source files, or understand the styling patterns in a frontend repo — even if they just say "what does this app look like?" or "pull out the design from this code.

Use when a change touches anything a user sees, such as components, pages, views, screens, styles, templates, dialogs, forms or copy, and before opening a PR for it. Covers driving the change in a real browser through every state, running a review sweep against the repo's design and framework guidelines, and capturing screenshots for the PR, including what to do when no browser is available.

>- Use when designing or changing Croct content models: regular vs union components, what becomes a component, slot, structure, or primitive attribute; writing component schema JSON (`structure`, `union`, `reference`, `text`, `number`, `boolean`, `list`, `@croct/file`, `@croct/richText` rich text, `format`, `choices`, `optional`, `private`, `position`); writing content definition JSON (`static`/`dynamic` values, `nullable`, `default`, `{{placeholder}}` paths); slot default, localized, experience, experiment, or content group content; schema limit errors (30 attributes, 6 nesting levels, 10 union members, complexity score 50); component versioning and breaking changes (`home-hero@2`); refactoring (duplicate and modify, duplicate and replace, flatten and merge); Markdown in text attributes with `@croct/md-lite`. Phrasings: "model content for a hero", "add an attribute to the banner component", "support variations", "component is too complex", "localize the hero", "translations", `preferredLocale`.

Use when starting a feature, refactor or bug fix; when asked to plan, spec, design or debug something; when writing a plan or handoff document; when choosing, adopting or upgrading a dependency, framework or external API; or when starting work that will become its own branch. Covers propose-before-building, scoping work to the PR size budget, debugging discipline, reading current docs instead of trusting memory, and when and how to use a git worktree.

Founden MCP server. Build your company over chat with one API key, no browser required. Works with Claude Desktop, Cursor, VS Code.

>- Convert frontend code (Vite, React, Angular, Vue, etc.) to a Stitch Design by chaining static HTML extraction, design system extraction, and file upload. **ALWAYS** use this skill when the user's intent is to move existing web apps or React/Angular/Vue components into Stitch (e.g., requests to "save", "migrate", or "upload"). You must use this skill even for simple "save" operations, as it is the only way to ensure the design system is extracted and assets are properly linked.

Function-first design guidance for UI, UX, product, information, and visual design. Use when designing, critiquing, or reviewing interfaces, screens, components, layouts, flows, charts, copy structure, APIs/CLIs as user surfaces, or any object meant to be used; when choosing between aesthetic options; when asked to "make it look better/cleaner/simpler/modern"; or when deciding whether to add, keep, or remove an element, animation, or feature.

Transforms vague UI ideas into polished, Stitch-optimized prompts. Enhances specificity, adds UI/UX keywords, injects design system context, and structures output for better generation results.

Analyze Stitch projects and synthesize a semantic design system into DESIGN.md files

Read this skill before the first step of any request to do something in an app on the person's own computer (Notes, Finder, System Settings, any desktop app), to look at their screen, or for \"computer use\". Computer use (desktop control) lets Claude take screenshots of the person's desktop and control it with clicks, typing and scrolling through the Claude desktop app; its tools are named mcp__computer-use__* when the session runs in the desktop app and mcp__remote-devices__computer_* when a cloud session is linked to the person's computer; before computer use is turned on for a conversation there may be no such tools, only an enable__mcp__remote-devices__computer tool, which turns it on. It covers turning it on, picking the right tool, the access flow, and the safety rules for tiered apps, links and financial actions. It is not for websites, which go through Claude in Chrome or the built-in browser and their own skills.

>- DESIGN.md 设计契约工作流——为任意项目生成 / 提炼 / 验收前端设计系统。 按 Google Stitch 规范的 DESIGN.md（YAML 令牌 + 11 节正文）沉淀设计决策， 产出机器可读 tokens、tokens.css（双形式令牌）、Tailwind 映射与可视化预览页， 并用 WCAG 2.1 对比度门禁做回归验收（可在 CI 中阻断构建）。 集成 awesome-design-md（84 个真实品牌 DESIGN.md 参考库）与 DeepSeek Search/Fetch 实时网页研究（付费直连、**默认关闭**——需用户显式同意并带 --yes-billing；无 key / 未同意时 优雅降级为内置抓取）。 TRIGGER when: 用户要给项目立设计规范/统一样式（"设计系统"、"DESIGN.md"、"design tokens"、 "配色规范"、"前端风格不一致"）；要从参考网站提炼配色/字体/圆角；要审计 UI 对比度 / 可访问性（WCAG、contrast）；要引用 awesome-design-md 里的现成 DESIGN.md； 为 UI 工作做 design research（竞品视觉分析、品牌参考）。 DO NOT TRIGGER when: 纯后端 / CLI / 数据管道任务；只需要生成一张图片（用 qianwen-image-generation）；只是普通代码功能开发且不涉及视觉一致性。