Art & Design

Use to establish or audit a product's brand so the factory's output looks intentional, not template — when a ticket asks to "create a brand", "define the design system / visual identity / palette / typography", or when delivered UI looks generic and needs a deliberate direction. Detects existing brand and extends it in its own idiom, or commits to one specific direction (palette, type pairing, voice) and writes a BRAND.md. Invoke whenever the factory needs a design foundation before building UI.

Design, build, improve, or audit websites and web interfaces; write or edit site copy, product instructions and technical documentation, including RU/EN translation. Use for landing pages, product UI, responsive components, redesigns and readiness checks. Scope text-only work, small fixes and audits to the request. Not for backend-only implementation or native apps.

Check whether a website can be found, read and quoted by AI search (ChatGPT, Claude, Perplexity, Google AI Overviews) and regular search. Audits robots.txt rules for every major AI crawler, firewall blocking, content that only appears with JavaScript, structured data, redirects, the sitemap and llms.txt, then reports fixes in priority order. Use when someone asks "can ChatGPT see my site?", "is my site blocking AI crawlers?", "why doesn't my site show up in AI answers?", for an AEO, GEO or AI vi

Before/after check of a web UI change at minimal image-token cost. It returns a text report of what changed (styles, text, layout, unexplained pixels) and at most one small cropped image. Use when verifying a frontend edit, checking for visual regressions, or comparing screenshots of a page before and after a change.

Multi-stage design & image critique-and-fix optimizer for graphic/brand & UI/UX screens; sibling of code-deep-optimizer/document-critique. Ingests a screenshot, URL, code (HTML/CSS), or spec and runs an 11-pass critique (hierarchy, gestalt, typography, color, usability heuristics/Laws of UX, WCAG, affective/trust, metrics, brand-parity, hallucination guard) over the frontend-ui spokes; severity-rates Blocker\u2192Nit and, for code-backed designs, applies Medium+ fixes to convergence (re-render/c

Use before building a view of stored data (card, panel, table, dashboard, report, profile page), and when a stakeholder says what to show, a field looks obviously useful, layouts for the same records compete, or several cards or charts cover the same categories. Not for pure looks (design-process) or how a posted chart reads (practical-charts).

Use when designing or restyling a UI (landing page, app screen, dashboard, component), on "make it look good", "less generic" or "more premium", when it looks AI-generated or bland, or for several visual directions, a design critique or a pre-ship polish pass. Not for Tailwind mechanics (tailwind).

When the user is planning, building, auditing, or growing an online tool website (免费工具站/工具站) whose primary growth channel is organic search — calculators, converters, generators, checkers, graders, lookup databases, single-purpose utilities. Also use when the user mentions '工具站', '在线工具', '免费工具', '计算器站', '转换器', '生成器', '小工具矩阵', '工具页', 'tool site', 'calculator site', 'converter site', 'free tools website', 'utility site', 'micro tool', 'tool directory', 'pSEO tool pages', 'programmatic tool pages',

Two effects, deliberately. Motion is where a site starts looking machine-made:

UI 코드와 디자인 토큰을 분석하여 디자인 품질 테스트를 자동 생성한다.

Website hero carousel: photos that take turns, and the tests that keep them honest

Owns the **how it looks and performs visually** layer: mobile graphics, dark

1. **코드 품질 평가 금지** — 아키텍처, 성능, 코드 스타일을 평가하지 마라. 디자인 원칙 준수 여부만 판정한다.

1. **컬러 변형만으로 시안 구분 금지 — 구별성은 선언이 아니라 계산이다** — 시안들은 서로 다른 레이아웃/구성 접근이어야 한다. 색상·토큰 값·카피 문구·아이콘 교체는 `axis_vector` 원소로 세지 않으므로, 그것만 다른 두 안은 Hamming distance 0 이며 FAIL 이다. 판정식과 실행 가능한 게이트: `../../references/visual-change-protocol.md` §5 Variant Contract Matrix. 실측 근거는 글로벌 REJECT `UI-04` — 계약이 4 축을 **이미 명시**했는데도 두 안이 전 축 동일값이었다.

1. **토큰 네이밍에 구체적 값 금지** — `blue-500`, `primary-blue` ✗ → `color.text.primary`, `color.background.brand` ✓. 값이 바뀌면 이름과 괴리가 생기고, 리브랜딩/다크모드 전환 때 즉시 무너진다.

Build and deploy world-class websites with any agent: briefing, direction, tokens, sections, components, content, motion, quality gates. Triggers on build me a website, landing page, portfolio, redesign, audit this site, deploy to higgsfield.

Reviews and improves web accessibility for keyboard users, screen readers, contrast, semantics, forms, and responsive interfaces. Use when the user asks for an accessibility audit, WCAG review, or wants to make a UI usable by more people. Trigger on "accessibility", "a11y", "WCAG", "доступность", "скринридер", "клавиатура", "контраст".

プログラミング言語やフレームワークの「初心者向け参考書（入門ガイド）」を、1 ファイルの HTML として作るスキル。TypeScript・Rust・Angular・Go の入門ガイドと同じデザイン（表紙・目次・キーワード検索・章末の確認問題・コードのコピー・読んだ章の記録・ダークモード）で、コード例はコンパイラや公式 Playground で実際に確かめ、エラーの波線・型のヒント・実行結果を載せる。「Python の参考書を作って」「React の入門ガイドがほしい」「〇〇の初心者向けの資料を HTML で」「beginner reference book for Kotlin」のように、言語やフレームワークの入門書・参考書・学習資料・チュートリアル冊子を作りたいときは、言い方が違っても必ずこのスキルを使う。/beginner-reference-book のあとに言語名が書かれていないとき、言語やフレームワーク以外のものが書かれたときの聞き返しも、このスキルが担当する。

为后台管理系统编排设计工作流程并制定或审查业务与组件交互规范。按理解、定义、方案、验证组织交付物与 AI 自检，参考 Cloudscape 判断功能取舍，综合 WAI-ARIA APG、Carbon、Ant Design、Radix 与 React Aria 规定组件行为，并提供设计评估、WCAG 2.2 AA 验收与设计系统令牌规范。适用于后台需求设计、流程规划、功能查漏、组件行为设计、AI 生成界面约束、交互评审与验收、设计系统规范制定，以及提炼或应用工作台、控制台参考风格（含 NextLabs、LarkAgentX 案例）。纯视觉美化、营销站或后端接口设计不单独触发。

Create polished, validated architecture, workflow, sequence, data-flow, and lifecycle/state diagrams as explorable standalone HTML with inline SVG, dark/light themes, optional trace motion, and PNG/JPEG/WebP/SVG/WebM export. Accept plain-language requirements or pasted Mermaid flowchart, sequenceDiagram, and stateDiagram input; inspect repository evidence when the diagram must reflect real code. Use when the user asks to visualize system architecture, infrastructure, cloud/security/network topol

Recursively turn generic or polished-but-forgettable landing pages, portfolios, and product-marketing surfaces into ownable design directions. Use when the user says a design is boring, generic, mediocre, shallow, templated, lacks taste or depth, asks for top-tier repeated refinement, or when a high-stakes design task needs divergent concepts before implementation. Orchestrates binary direction branching, visual evidence, pairwise critique, backtracking, and staged implementation; treats design-

Maps a reference (site, screenshot, Figma, CSS, brand book) to OpenDesigner dials, or writes its Taste Index. Copies structure and quality, never another brand's identity.

Generate user personas from research data, create journey maps, plan usability tests, and synthesize research findings into actionable design recommendations.

Design systems and the pages built on them, one decision at a time. Use for a landing page, an app or a redesign ("make it look professional"), a style direction, a taste poll, tokens or DESIGN.md.