qiaomu-design

作者 joeseesun已验证

偏执型设计顾问:反 AI 味设计 + 风格试衣间 + 58 站设计系统库的 Claude Code Skill | Opinionated design advisor for Claude Code: anti-generic UI, style fitting room, 58 real-site design systems.

503
Stars
38
Forks
JavaScript
语言
2026/8/23
添加时间

⚠️ 第三方软件声明

本 Skill 为第三方开源软件,独立托管于 GitHub。SkillTip 仅为信息目录,不控制或维护底层仓库。所显示的安全检查为自动化且范围有限,安装前请自行审查源码。

阅读服务条款

安装

添加到你的 Claude Code skills 目录:

# Add to your Claude Code skills
git clone https://github.com/joeseesun/qiaomu-design

快速入门

使用 qiaomu-design 等 Skills 的指南。

安全报告

已验证

上次扫描:—

{
  "status": "PASSED",
  "issues": []
}

README.md

qiaomu-design · 偏执型设计顾问

中文 | English

让 AI 做出"不像 AI 做的"设计:Jobs 式产品直觉 + Rams 式功能纯粹主义,融合五大顶级设计 Skill 的实测精华。

An opinionated design advisor skill for Claude Code: anti-generic aesthetics, engineering-grade delivery, and a visual "style fitting room" — distilled from a controlled experiment across 5 top design skills.

License: MIT Skill 更新日志

已验证: 本 skill 的每条核心规则均来自多轮主流设计 Skill 横评实验,并经浏览器实测交互验证。

完整横评: 前端设计 Skill 横评实验室 展示 8 个主流变体 × 10 个任务的 80 个真实生成页面,qiaomu-design 与 impeccable 均已补齐全部任务。

模型建议

强烈建议用 GLM5.2 / GLM 5.2Claude 模型执行本 skill。近期实测里,Codex 中的 GPT 模型做设计生成效果很不理想:容易保守、空洞、模板化,难以稳定产出有气质的完整页面。追求高质量视觉时,优先切到 GLM5.2 或 Claude。

案例

视频案例 · 冥想网站

qiaomu-design 生成的冥想网站案例动图

横评截图

以下 10 张截图来自 qiaomu-design 在横评实验里的真实生成页面。点击截图可打开线上可交互原始文件。

落地页 · AI 会议笔记仪表盘 · 电商运营
qiaomu-design 生成的 AI 会议笔记落地页截图qiaomu-design 生成的电商运营仪表盘截图
作品集 · 独立设计师交互向导 · 机票预订
qiaomu-design 生成的独立设计师作品集截图qiaomu-design 生成的机票预订交互向导截图
设置中心 · 危险操作组件创意 404 · 星空航行
qiaomu-design 生成的设置中心组件面板截图qiaomu-design 生成的创意 404 页面截图
多样性挑战 · 三向播放器电商详情 · 手冲咖啡器
qiaomu-design 生成的三向音乐播放器设计截图qiaomu-design 生成的手冲咖啡器电商详情页截图
移动端 · 冥想助眠 App数据叙事 · 精品咖啡十年
qiaomu-design 生成的冥想助眠移动端页面截图qiaomu-design 生成的精品咖啡数据叙事页面截图

这是什么

一个 Claude Code Agent Skill。安装后,当你说"帮我设计 / 重新设计 / 优化界面"时,它会接管设计流程:先诊断真实需求,再用可视化风格试衣间给你 4 个方向实际看着选,确认后按工程验收标准交付——而不是直接吐一个紫渐变居中 Hero 的"AI 味"页面。

为什么值得用

大多数 AI 设计的问题不是画不好,而是:默认审美收敛(Inter + 紫渐变 + 三等分卡片)、方向靠文字描述拍脑袋选、交付物过不了工程验收、中文排版直接套英文规则。本 skill 针对这四个问题各给了一套实测过的机制。

核心能力

能力你得到什么
风格试衣间4 个互斥方向的真实迷你 mockup,生成在 design-previews/YYYY-MM-DD-任务名/index.html,点选/键盘进入确认弹层,带拨盘与建议一起回传
设计读取 + 三拨盘按任务类型自适应冒险度/动效/密度:功能页收敛、开放命题放开
AI 反套路禁令禁 AI 紫渐变、禁 Inter、禁斜体、禁居中套路、禁 AI 文案词——从源头消灭"AI 味"
中文排版规范系统字体栈优先、装饰中文字体子集化、盘古之白、行高/字重/标点纪律
Emil 动效工艺套件emil-design-eng + review-animations + animation-vocabulary + apple-design:写动效、审动效、命名动效、做流体手势
工程验收清单Vercel 规范:a11y、表单、焦点陷阱、危险操作防护,file:line 格式审查
Carbon 决策体系基于官方 319 页索引,整合基础 token、组件选型、完整工作流、数据可视化与无障碍;借方法,不复制 IBM 外观
58 站设计系统库Stripe/Linear/Apple 等真实网站 DESIGN.md(Google Stitch 格式),可直接"参考 XX 做"
打磨模式已有页面不推倒重来:Audit/Critique/Polish/Animate/Harden/Live 六动作
交付门禁preflight 强制检查清单,任何一条不过就不交付
自进化机制用户反馈自动抽象成规则写入偏好账本,越用越贴合你

快速开始

npx skills add joeseesun/qiaomu-design

然后在 Claude Code 里直接说:

帮我设计一个产品落地页
手动安装
git clone https://github.com/joeseesun/qiaomu-design.git
cp -r qiaomu-design ~/.claude/skills/qiaomu-design

前置条件:已安装 Claude Code

使用方式

触发词:重新设计 / redesign / 优化界面 / 设计方案 / UI 审查 / 帮我看看设计 / 参考 XX 的设计 / 给我一个设计系统

典型流程(三阶段,每阶段等你确认):

你:帮我重新设计博客首页
 ↓ Phase 1  诊断:设计读取 + 三拨盘 + 2-3 个关键问题
 ↓ Phase 2  风格试衣间:生成 design-previews/YYYY-MM-DD-任务名/index.html
            4 个方向实际看着选,确认后把方向、拨盘和建议回传
 ↓ Phase 3  执行:先立 DESIGN.md 锚 → 写码 → preflight 门禁 → 交付

已有页面要优化:说"帮我打磨这个页面 / 反 AI 味 / 加动效",走打磨模式,不推倒重来。

工作原理

SKILL.md                       人格 + 工作流 + 拨盘 + 反套路禁令 + 自进化协议
references/
  user-preferences.md          用户偏好账本(自进化写入,最高优先级)
  style-preview.md             风格试衣间规范(4 方向 + 手动确认回传)
  divergence-playbook.md       发散手册(轴级差异检验 + 14 种美学方向)
  motion-craft.md              动效工艺(Emil Kowalski 体系)
  motion-review.md             动效专项审查(Before/After/Why + 阻断标准)
  animation-vocabulary.md      动效术语词典(把模糊描述转成准确 motion brief)
  apple-fluid-interfaces.md    Apple 流体交互(手势、弹簧、速度交接、透明材质)
  engineering-checklist.md     工程验收(Vercel WIG + 组件行为标准)
  carbon-foundations.md        Carbon 网格/token/排版/动效/内容/a11y/AI
  carbon-components.md         Carbon 组件选型矩阵 + 七问 dossier
  carbon-patterns.md           Carbon 表单/对话/状态/CRUD 完整工作流
  carbon-data-visualization.md Carbon 图表选型/轴/颜色/图例/仪表盘
  carbon-source-map.md         319 页 sitemap + 321 MDX 覆盖与刷新方法
  chinese-typography.md        中文排版与配色(W3C clreq / Ant Design / Apple HIG)
  preflight.md                 交付前门禁
  design-systems/{58 站}/      DESIGN.md 设计系统参考库

实测验证

规则不是拍脑袋写的。改造前先做了一场受控实验:6 个变体(anthropics/frontend-design、vercel/web-design-guidelines、ui-ux-pro-max、taste-skill、emil-design-eng、无 Skill 对照组)× 7 个任务(落地页/仪表盘/作品集/交互向导/组件面板/创意 404/多样性挑战)= 42 个页面,横评视觉个性、工程规范、动效工艺、交互完成度、组件合理性、创造性、多样性七个维度。

后续复测扩展为 8 个变体 × 10 个任务的公开横评,覆盖电商详情、移动端与数据叙事等新场景。每个维度胜者的核心机制被移植进本 skill,来源在 SKILL.md「血统说明」逐条可查。

限制与边界

  • 这是 Claude Code 的 Agent Skill,不是独立软件;效果依赖模型能力
  • 风格试衣间默认生成浅层任务目录:design-previews/YYYY-MM-DD-任务名/index.html; 当前不在项目中时退到桌面目录;用户可以指定输出目录
  • 点选回传依赖本地预览服务;服务失败时降级为打开 HTML 文件并在对话中回复选择
  • 本 skill 只做设计,不会默认给生成页面注入打赏、公众号、GitHub/X 浮条或乔木个人 Profile; 对外页面如需署名,建议只放低干扰页脚 Powered by 向阳乔木 链接到 https://qiaomu.ai/
  • 58 站 DESIGN.md 库来自公开网站的设计系统提炼,用于风格参考,不代表对应公司背书
  • 装饰性中文字体默认禁用(体积 5-20MB),只在创意标题场景子集化加载——这是特性不是缺陷

来源与致谢

MIT License。融合机制来源(详见 SKILL.md 血统说明):anthropics/skills · vercel-labs/web-interface-guidelines · Leonxlnx/taste-skill · emilkowalski/skills · nextlevelbuilder/ui-ux-pro-max-skill · pbakaus/impeccable · arvindrk/extract-design-system · mattpocock/skills · IBM Carbon Design System(官方站点与 Apache-2.0 carbon-website 文档的重写摘要) · DESIGN.md 库基于 VoltAgent/awesome-design-md(Google Stitch 格式)

版权

MIT License。Copyright (c) 向阳乔木。

项目主页:qiaomu.ai · GitHub:@joeseesun


English

qiaomu-design is a Claude Code Agent Skill that makes AI-generated interfaces stop looking AI-generated. It fuses the experimentally-validated strengths of five top design skills (Anthropic frontend-design, Vercel web-interface-guidelines, taste-skill, Emil Kowalski's design engineering, ui-ux-pro-max) into one opinionated design advisor.

Install

npx skills add joeseesun/qiaomu-design

Then just ask Claude Code: redesign my landing page.

What you get

  • Recommended models: use GLM 5.2 or Claude for this skill. In recent runs, GPT models inside Codex produced poor design results: too conservative, sparse, and template-like for high-quality visual work.
  • Style fitting room: 4 mutually-divergent direction mockups (real fonts/colors/layout) in design-previews/YYYY-MM-DD-task/index.html. Pick by click or keys, confirm with dial values and optional notes, then the local preview server reports the selection back to the current workflow.
  • Design read + three dials: VARIANCE / MOTION / DENSITY auto-tuned per task type — restrained on functional UI, bold on open creative briefs.
  • Anti-slop bans: no AI-purple gradients, no Inter, no italics, no centered-hero clichés, no "revolutionary/seamless" copy.
  • Chinese typography rules: system font stack first, subset decorative CJK webfonts (5-20 MB otherwise), CJK spacing/punctuation/line-height discipline.
  • Motion craft (Emil Kowalski system) and engineering checklist (Vercel WIG): easing/durations/stagger, a11y, focus traps, destructive-action guards.
  • 58 real-site DESIGN.md library (Stripe, Linear, Apple…, Google Stitch format) for "make it like X" requests.
  • Polish mode for existing pages (Audit/Critique/Polish/Animate/Harden/Live) — no rewrites from scratch.
  • Pre-flight gate: a hard checklist; nothing ships if any item fails.
  • Self-evolution: your feedback gets abstracted into a preferences ledger the skill reads before every task.

Verified

Every core rule traces back to controlled design-skill comparisons. The first run covered 6 variants × 7 tasks × 42 generated pages; the public follow-up expands to 8 variants × 10 tasks × 80 generated pages. The qiaomu-design screenshots above link to the live interactive case pages. The meditation website case is embedded as an animated GIF above, with the source MP4 kept at docs/assets/cases/qiaomu-design-meditation-site.mp4. Provenance is documented per-module in SKILL.md.

Limits

Requires Claude Code. The fitting room is a local HTML preview served by a tiny local callback server; if the server cannot start, it falls back to opening the HTML file and asking the user to reply with the selected direction. qiaomu-design does not inject Qiaomu profile widgets into generated pages by default. The DESIGN.md library is distilled from public sites for reference, not endorsement.

MIT © 向阳乔木 · qiaomu.ai · @joeseesun

常见问题

What is qiaomu-design?

qiaomu-design is an open-source ai agents skill for AI coding assistants such as Claude Code, Codex CLI, and ChatGPT, built by joeseesun. 偏执型设计顾问:反 AI 味设计 + 风格试衣间 + 58 站设计系统库的 Claude Code Skill | Opinionated design advisor for Claude Code: anti-generic UI, style fitting room, 58 real-site design systems. It has 503 GitHub stars.

Is qiaomu-design safe to use?

Yes. qiaomu-design passed SkillsLLM's automated security scan — a dependency vulnerability audit plus prompt-injection heuristics — with no high-severity issues. You can read the full report in the Security Report section on this page.

How do I install qiaomu-design?

Clone the repository with "git clone https://github.com/joeseesun/qiaomu-design" and add it to your Claude Code skills directory (see the Installation section above). qiaomu-design ships a SKILL.md manifest, so compatible agents can discover and load it automatically.

What programming language is qiaomu-design written in?

qiaomu-design is primarily written in JavaScript. It is open-source under joeseesun on GitHub, so you can review or fork the full source.

Are there alternatives to qiaomu-design?

Yes. SkillsLLM lists many other AI Agents skills you can browse and compare side by side. Open the AI Agents category from the badge at the top of this page, or use the Related Skills and comparison links further down to weigh qiaomu-design against similar tools.

评论 (0)

暂无评论,成为第一个分享想法的人!

ECC

by affaan-m

10

The agent harness performance optimization system. Skills, instincts, memory, security, and research-first development for Claude Code, Codex, Opencode, Cursor and beyond.

242,21936,702JavaScript
AI 智能体ai-agentsanthropicclaude-code
查看详情
15

An agentic skills framework & software development methodology that works.

234,96620,863Shell
AI 智能体ai-agentsbrainstorming
查看详情

hermes-agent

by NousResearch

10

The agent that grows with you

234,43747,175Python
AI 智能体ai-agentsagent-orchestration
查看详情

The agent harness performance optimization system. Skills, instincts, memory, security, and research-first development for Claude Code, Codex, Opencode, Cursor and beyond.

185,94028,768JavaScript
AI 智能体ai-agentsanthropicclaude-code
查看详情

cc-switch

by farion1231

3

A cross-platform desktop All-in-One assistant for Claude Code, Codex, OpenCode, OpenClaw, Grok Build & Hermes Agent. Only official website: ccswitch.io

128,8688,826Rust
AI 智能体claude-codeai-tools
查看详情

claude-code

by anthropics

Claude Code is an agentic coding tool that lives in your terminal, understands your codebase, and helps you code faster by executing routine tasks, explaining complex code, and handling git workflows - all through natural language commands.

120,03119,897Shell
AI 智能体
查看详情

开发者还喜欢

基于喜欢此 Skill 的开发者投票和收藏

ECC

by affaan-m

10

The agent harness performance optimization system. Skills, instincts, memory, security, and research-first development for Claude Code, Codex, Opencode, Cursor and beyond.

242,21936,702JavaScript
AI 智能体ai-agentsanthropicclaude-code
查看详情
15

An agentic skills framework & software development methodology that works.

234,96620,863Shell
AI 智能体ai-agentsbrainstorming
查看详情

hermes-agent

by NousResearch

10

The agent that grows with you

234,43747,175Python
AI 智能体ai-agentsagent-orchestration
查看详情

n8n

by n8n-io

12

Fair-code workflow automation platform with native AI capabilities. Combine visual building with custom code, self-host or cloud, 400+ integrations.

201,88160,308TypeScript
MCP 服务器apisai-tools
查看详情

The agent harness performance optimization system. Skills, instincts, memory, security, and research-first development for Claude Code, Codex, Opencode, Cursor and beyond.

185,94028,768JavaScript
AI 智能体ai-agentsanthropicclaude-code
查看详情

cc-switch

by farion1231

3

A cross-platform desktop All-in-One assistant for Claude Code, Codex, OpenCode, OpenClaw, Grok Build & Hermes Agent. Only official website: ccswitch.io

128,8688,826Rust
AI 智能体claude-codeai-tools
查看详情