Agent skill

Web Design

by xiaopu-ai in xiaopu-ai/web-design

Web 视觉设计 SKILL。输入 PRD / 参考 URL / 截图 / 关键词(任意组合),先产出一份标准化 DESIGN.md 设计规范,用户确认后据此生成 UI/UX、视觉、动效、响应式全部达标的 web 代码。专攻 web 端:Landing Page、Portfolio、产品页、博客、个人站、SaaS 介绍页等。当用户说"帮我做个网站""设计一个页面""参考 XX…

MITAuto-check passedFrontend & Design

Install Web Design

skills CLI
$ npx skills add xiaopu-ai/web-design --skill web-design -a claude-code

Project install by default; add -g for ~/.claude/skills/.

GitHub CLI
$ gh skill install xiaopu-ai/web-design web-design --agent claude-code

Project scope by default; add --scope user for a personal install. Needs GitHub CLI 2.90.0 or later (public preview).

Claude Code skills documentation · loads skills from .claude/skills/

Facts

Skill name
web-design
GitHub stars
783
Token cost
~3k tokens
SKILL.md length
1,067 words
Files
127 (incl. scripts, references)
Skills in repo
1
Repo updated
First seen
Licence
MIT

At a glance

Web 视觉设计 SKILL。输入 PRD / 参考 URL / 截图 / 关键词(任意组合),先产出一份标准化 DESIGN.md 设计规范,用户确认后据此生成 UI/UX、视觉、动效、响应式全部达标的 web 代码。专攻 web 端:Landing Page、Portfolio、产品页、博客、个人站、SaaS 介绍页等。当用户说"帮我做个网站""设计一个页面""参考 XX…

  • Works in 3 steps: 项目里已经有 DESIGN.md 了吗?→ 有则沿用/修改/重建 → 项目里有 PRD.md 吗?→ 有则进入 A4(PRD 驱动流程) → 都没有 → 正常走 Phase A
  • Tasks that involve PRD writing
  • SKILL.md covers 核心流程, Phase A: 理解需求, Phase B: 输出 DESIGN.md(必走) and Phase C: 生成项目代码(必走), plus 5 more sections
  • Runs JavaScript scripts from its folder; calls python3 and curl

What it does

Web Design is an agent skill from xiaopu-ai/web-design. Web 视觉设计 SKILL。输入 PRD / 参考 URL / 截图 / 关键词(任意组合),先产出一份标准化 DESIGN.md 设计规范,用户确认后据此生成 UI/UX、视觉、动效、响应式全部达标的 web 代码。专攻 web 端:Landing Page、Portfolio、产品页、博客、个人站、SaaS 介绍页等。当用户说"帮我做个网站""设计一个页面""参考 XX 做一个""把这个截图/PRD 做成网页""做一个 landing page""出一份 design 规范"时触发。不用于后端、数据库、纯逻辑 bug 修复。

Its SKILL.md is about 3k tokens, which your agent loads only when the skill is triggered. The skill folder holds 129 other files, including scripts and reference files (for example `README.md`, `docs/DESIGN.md` and `docs/app.js`).

It sits in Frontend & Design, covering PRD writing, Design tokens and Landing pages. It works with Playwright. The repository describes itself as: A Claude Code SKILL for designing beautiful, consistent web pages — spec first, code second. The licence is MIT.

When your agent uses it

  • Tasks that involve PRD writing
  • Tasks that involve Design tokens
  • Tasks that involve Landing pages

Example prompts

  • “帮我做个网站”
  • “设计一个页面”
  • “参考 XX 做一个”
  • “/web-design”

Requirements

  • Python 3
  • Node.js

Workflow steps

3 steps, taken from the first numbered list in SKILL.md.

  1. 项目里已经有 DESIGN.md 了吗?→ 有则沿用/修改/重建
  2. 项目里有 PRD.md 吗?→ 有则进入 A4(PRD 驱动流程)
  3. 都没有 → 正常走 Phase A

What it can do on your machine

Read from SKILL.md and the folder at commit 22a4f48. It shows what the files ask for, not the result of running them.

  • Tool permissions

    Pre-approves nothing: there is no allowed-tools line, so your agent's usual permission prompts apply.

    From allowed-tools in the SKILL.md frontmatter.

  • Runs code

    Ships 1 file in scripts/ (JavaScript, from the files we listed), which the agent can run.

    Shell commands in SKILL.md call:

    • python3
    • curl

    From the folder's file list and the shell code blocks in SKILL.md.

  • Network

    Links to these hosts (documentation or services it may open):

    • github.com

    From URLs in SKILL.md, links to its own repository left out.

  • Credentials

    Names no API keys, tokens, secrets or passwords.

    From names ending in _API_KEY, _TOKEN, _SECRET, _KEY or _PASSWORD in SKILL.md.

Context cost

Web Design loads about 3k tokens when it runs, and up to ~249k if it reads all its reference files. Until then it costs about 70 tokens; SKILL.md has 1,067 words of instructions outside code blocks.

Always · name and description, kept in context so the agent knows when to use it
~70
When it runs · the whole SKILL.md, loaded when a task matches
~3k
With references · SKILL.md plus every file in references/, read only if the agent opens them
~249k

Estimates: characters ÷ 4, the usual rule of thumb; real counts depend on the model's tokenizer. Scripts and assets cost tokens only if the agent reads them.

Safety

Auto-check passed

The automated check found no risky patterns in SKILL.md.

Automated static check — not a guarantee. Review scripts before installing. It scans the text of SKILL.md for risky patterns (piping downloads into a shell, reading credential files, hidden Unicode, destructive commands); the scripts in this folder are not scanned.

SKILL.md

The full file from xiaopu-ai/web-design at commit 22a4f48, republished under its MIT licence (© xiaopu-ai). 1,067 words, ~2,951 tokens.

Download SKILL.mdSave it as .claude/skills/web-design/SKILL.md (or your agent's skills folder). This skill also uses 126 other files; get the full folder from GitHub.
name
web-design
description
Web 视觉设计 SKILL。输入 PRD / 参考 URL / 截图 / 关键词(任意组合),先产出一份标准化 DESIGN.md 设计规范,用户确认后据此生成 UI/UX、视觉、动效、响应式全部达标的 web 代码。专攻 web 端:Landing Page、Portfolio、产品页、博客、个人站、SaaS 介绍页等。当用户说"帮我做个网站""设计一个页面""参考 XX 做一个""把这个截图/PRD 做成网页""做一个 landing page""出一份 design 规范"时触发。不用于后端、数据库、纯逻辑 bug 修复。

Web Design

Web 端视觉设计 SKILL。两阶段工作流:先出规范(DESIGN.md),再出代码。

DESIGN.md 是显式文件产物,保存到项目目录,可跨项目复用、可手动修改、可给其他工具消费。

核心流程

Phase A  理解需求(灵活输入)──→  Phase B  输出 DESIGN.md(用户确认)──→  Phase C  生成项目代码

启动时先检测(按优先级):

  1. 项目里已经有 DESIGN.md 了吗?→ 有则沿用/修改/重建
  2. 项目里有 PRD.md 吗?→ 有则进入 A4(PRD 驱动流程)
  3. 都没有 → 正常走 Phase A

Phase A: 理解需求

输入是灵活的,以下任意组合都行,不要强制用户走固定路径:

输入处理方式
参考 URL抓取 HTML+CSS,提取 Token,motion audit
截图 / 设计稿从视觉提取气质、色温、密度、字体风格
关键词 / 描述"暗色克制衬线风格" → 从 style-seeds 匹配 Token
品牌名"做成类似 Linear 的" → 先查 references/design-systems/,有则直接读取;无则抓取该网站
混合"参考这个 URL 但配色要暖一些" → 提取 + 覆盖
PRD 文档读取 PRD.md 的产品定位、页面结构、设计交接区 → 分析竞品风格 → 推导设计方向
A1. URL 分析(有参考 URL 时)

意图判断:参考(提取设计语言,做独立新页面)vs 复刻(用户明确说"复刻/clone")。

抓取流程(按优先级尝试):

  1. 首选:Playwright 爬虫(真实浏览器,能绕反爬、渲染 SPA、模拟滚动)

    bash
    python3 scripts/crawl_website.py --url [URL] --output ./crawl-output --scroll-delay 600

    输出:每屏视口截图 + tokens.json + structure.json + styles.css

  2. 备选:Token 提取脚本(轻量,适合简单静态站)

    bash
    python3 scripts/extract_design_tokens.py --url [URL] --format json
  3. 降级:手动 curl(脚本都不可用时)

    bash
    curl -Ls [URL]

综合三类输入判断:实际体验(氛围节奏)> 截图(局部气质)> Token 数值(精确参数)。Playwright 爬虫的视口截图可直接作为"体验"和"截图"两类输入。

Motion Audit(参考站有滚动联动动效时必做):

  • 触发信号:动效跟随滚动连续变化 / 模块间有 stagger-pin-parallax / 引用 GSAP-Lenis 等
  • 记录每个动效区块的 trigger、driver、elements、effect、timing

降级链路:

  • Playwright 爬虫失败 → 查 references/design-systems/ 有无该品牌预置文件(58 个品牌)
  • 预置文件也无 → extract_design_tokens.py 尝试静态提取
  • 静态提取也失败 → 手动 curl CSS
  • 全部失败 → 要求用户提供截图,从视觉手动建 Token
A2. 对话引导(无参考物时)

快速锁定风格方向,不要机械走问答清单:

核心问题(按需选问,不全问):

  • 亮色 vs 暗色?
  • 衬线 vs 无衬线?
  • 克制留白 vs 丰富饱满?
  • 强调色偏好?
  • 有没有喜欢的网站或风格参考?

从 references/style-seeds.md 匹配最近似的种子(10 种预设方向),展示给用户确认或混搭。

用户提到具体品牌时,查阅 references/design-systems/INDEX.md 按类别检索,按需读取对应品牌文件(58 个真实网站设计规范,每个 ~300 行)。品牌文件只含静态设计系统,动效仍需按交互档位补充。

A3. 交互档位确认(不可跳过)
档位体验感
L1 精致静态优雅 hover + 柔和入场
L2 流畅交互滚动 reveal、视差、导航变化
L3 沉浸体验滚动驱动时间线、pin、光标跟随、转场

L2 追问:reveal 风格(fadeInUp/scaleIn)、视差、导航变化 L3 追问:section pin、光标效果、转场、是否允许 GSAP/ScrollTrigger/Lenis

查阅 references/interaction-patterns.md 获取对应档位的完整代码。

A4. PRD 驱动(检测到项目级规范文件时)

启动时自动扫描项目根目录,按优先级:

  1. PRD.md / prd.md
  2. SPEC.md / spec.md
  3. README.md 或任何 .md — 若含「产品定位 / Target Users / Pages / 核心页面」等关键字段

不绑定特定工具或模板。只要能从文档里抽到下列字段,就足以作为设计输入:

字段(中/英别名)用途
产品名称 / Product / NameHero 文案
一句话定位 / Tagline / PitchHero 副标题、定调
目标用户 / Target Users / Audience色彩温度、字体风格决策
核心页面 / Pages / ScreensPhase C 生成清单
竞品 / Competitors / ReferencesPhase A 风格参考起点
技术栈 / Tech StackPhase C 代码生成方案
设计交接 / Design Notes / 调性直接继承的硬约束

竞品风格分析:

  • PRD 提到竞品 → 优先查 references/design-systems/ 是否已有预置
  • 有预置 → 读取该品牌规范作为起点
  • 无预置 → 调用 scripts/crawl_website.py 抓取提取
  • PRD 没提 → 按产品类型 + 调性从 references/style-seeds.md 匹配最近种子

提取不到关键字段时:不要瞎猜,回到对话引导(A2)补齐。

与用户确认(示例):

我从 {文件名} 读到了:

  • 产品:{名称} — {一句话定位}
  • 目标用户:{画像}
  • 参考竞品:{列表 或 "无"}
  • 建议的风格方向:{1-2 个候选,各附一句理由}

这个方向对吗?

确认后进入 A3(交互档位),然后 Phase B。


Phase B: 输出 DESIGN.md(必走)

必须生成 DESIGN.md 文件。 按 references/design-md-template.md 模板输出 9 个章节:

  1. Visual Theme & Atmosphere — 设计哲学、氛围关键词、一句话定调
  2. Color Palette & Roles — 完整 CSS 变量定义(含 RGB 辅助值)
  3. Typography Rules — 字体族、Google Fonts URL、字号层级表、禁止字体
  4. Component Stylings — 按钮/卡片/导航/链接/标签等完整 CSS(含所有状态)
  5. Layout Principles — 网格、间距梯度、容器宽度
  6. Depth & Elevation — 阴影体系
  7. Animation & Interaction — 动效档位、入场/滚动/悬停/特效的完整代码
  8. Do's and Don'ts — 设计护栏、Anti-Patterns(至少 8 条)
  9. Responsive Behavior — 断点、触摸目标、折叠策略

文字装饰规则:生成 Color Palette 和 Typography 后,按 references/text-decoration-rules.md 决策表对标题逐级判断是否添加渐变/投影。

展示摘要给用户确认,确认后保存 DESIGN.md 到项目目录,再进入 Phase C。


Phase C: 生成项目代码(必走)

C1. 确认场景 + 收集内容

问用户做什么页面,收集具体内容。查阅 references/scene-defaults.md 获取该场景的布局和组件基线。

常见场景的内容需求:

场景需要收集
Landing Page标题、副标题、Feature 列表、CTA、Social proof
个人站 / Portfolio名字、头衔、项目列表、关于、联系方式、博客列表
Blog名称、文章列表/内容、作者信息、分类标签
产品页产品名、卖点、功能截图、定价、FAQ

用户信息不全时主动问;用户说"你先做,内容我后面填"则用合理的占位内容。

C2. 设计框架 + 生成代码
  1. 检查项目环境:框架、路由、样式方案、现有组件和 assets
  2. 框架决策:有项目 → 沿用;无项目 → 最小可行方案
  3. 严格按 DESIGN.md 生成:
    • 所有颜色通过 CSS 变量引用
    • 字体按 DESIGN.md 定义使用
    • 交互按 DESIGN.md 指定档位实现
    • 不违反 Do's and Don'ts
  4. 图片策略:用户素材 > 项目 assets > 参考站 URL 占位 > Unsplash
  5. 图标策略:项目现有库 > lucide-react > 内联 SVG(参考 references/icon-library.md)
C3. 审计
  • DESIGN.md 合规检查:代码是否严格遵循规范
  • 参考物差异审计(有参考时):Token / 版式 / 字体 / 交互 / 素材逐项比对
  • 响应式验证:至少移动端 + 桌面端
  • 质量清单:读 references/quality-checklist.md 逐项自检

首页爆点原则(Landing Page 专属)

一个 Web 平台最重要的是落地页(首页)。其他页面保持规范、风格统一即可,唯有首页必须满足下面两条硬性原则,不可妥协。

原则一:3 个「卧槽」爆点 + 1 个巧思

爆点位置固定,每个位置必须有一个让用户"卧槽"的视觉冲击:

#位置要求典型招式
1首屏 / Hero加载瞬间就被动态/3D/巨型渐变震住卡片星座爆发、WebGL 3D 物体、巨型标题 mask reveal、鼠标跟随光斑、3D 卡片堆
2首次滑动离开首屏的第一个 section 必须有"钩子"继续吸引宣言横滚带(大字 marquee)、数字爆发、分屏切割、卡片汇聚转场、全屏视频
3列表 / 阵列 / 枚举 / 展示区不能是普通等大 gridBento 不等大布局、SpotlightCard hover 跟踪光、3D 倾斜卡片、抽象艺术顶

1 个巧思设计点:一处"不注意就错过、一注意到就笑"的细节。候选:

  • 交互反馈微动画(复制后的 sparkle + tooltip)
  • 彩蛋(Konami / Ctrl+K 命令面板)
  • Meta 梗(关键词 hover 弹 "这就是本页用到的规范 →")
  • 滚动到底的小惊喜(如"你看到这里了,谢谢")
原则二:流畅 · 滑动无阻(不可破)

以下是性能红线,与爆点设计同等优先级。爆点炸裂但卡顿 = 零分。

  • 3D / WebGL:全页最多 1 处,且必须 IntersectionObserver 不可见时暂停
  • filter: blur() on moving elements:禁止。用 opacity + scale 做景深代替
  • backdrop-filter: blur():值 ≤ 14px,且不覆盖大面积滚动区
  • Lenis / scroll-jacking:仅当 pin-scrub 真正必要时启用;普通页面用原生 scroll-behavior: smooth
  • ScrollTrigger pin:单页 ≤ 2 处(doubao 级叙事才解锁)
  • pointermove 监听:必须 rAF 节流
  • Custom cursor 全局替换:仅在气质强烈的设计类/工具类站点用,普通 SaaS 禁用
实现建议(性能友好的爆点配方)

这些组合实测在中等配置设备可达 60fps:

爆点实现性能成本
鼠标聚光灯CSS var --mx/--my + radial-gradient(at var(--mx) var(--my))⭐ 极低(单次 repaint)
卡片爆发入场GSAP 一次性 timeline(load 时 stagger 1.5s,结束即释放)⭐⭐ 低(仅 1 次)
宣言横滚带纯 CSS transform: translateX keyframes⭐ 极低
SpotlightCard每张卡 --mx/--my + ::before radial-gradient(rAF 节流)⭐⭐ 低
大字 mask reveal纯 CSS clip-path 一次性动画⭐ 极低
关键字渐变流动background-position 动画⭐ 极低

避免的反面组合:

  • ❌ 12+ 元素同时带 filter: blur() + will-change(GPU 内存爆)
  • ❌ Lenis + 多个 pin-scrub + WebGL(主线程堵死)
  • ❌ Three.js 常驻渲染不暂停(即使离开视窗也吃 FPS)

100 分质量底线(硬性验收)

DESIGN.md + 生成代码必须同时满足以下全部条件,任何一条不满足都算未完成。完整细项见 references/quality-checklist.md,这里是最核心的红线:

DESIGN.md 红线
  • 9 个章节都有实质内容,不是模板占位
  • 每个组件样式包含 default / hover / active / focus / disabled 全部状态
  • 所有颜色定义为 CSS 变量,含 RGB 辅助值(便于 rgba)
  • 字体有 @import URL + fallback stack
  • 动效档位明确(L1/L2/L3),对应依赖声明完整
  • Do's and Don'ts ≥ 8 条,其中 Don'ts ≥ 5 条
  • L2+ 必须包含 prefers-reduced-motion 降级
  • 响应式至少覆盖 Desktop + Mobile
Show full SKILL.md (440 more words)Show less
代码红线
  • 所有颜色通过 CSS 变量引用,零硬编码 hex
  • 字体严格按 DESIGN.md 的 Typography Rules 实现
  • 每个可交互元素有 hover + focus 态
  • 入场动画已实现(至少 L1 级 fadeIn)
  • L2+ 实现滚动 reveal + 导航滚动态 + 视差
  • L3 实现 pin / 光标跟随 / 转场
  • 图片禁用纯色块占位(参考站 URL / Unsplash / 用户素材)
  • 图标用项目库 / lucide-react / 内联 SVG(Playful 调性外禁 Emoji)
  • 未违反 DESIGN.md Do's and Don'ts 中的任何一条
  • 移动端 ≤ 600px 无横向溢出,触摸目标 ≥ 44×44px
签名动效红线(L2+ 强制,L3 加倍)

任何 L2 及以上页面必须包含以下 6 类动效,缺一不可。优先从 vue-bits / reactbits 目录挑(查 references/motion-library.md),能复用源码就不要自己写。

类别数量落点
Text Animation — Hero H1≥ 1大标题入场或持续态(如 SplitText / ShinyText / GradientText)
Text Animation — Section H2≥ 1章节标题滚动触发(如 ScrollFloat / BlurText)
Text Animation — Body / Label≥ 1正文、eyebrow、代码块(如 ScrollReveal / TextType / ScrambleText)
Animation — 元素级≥ 1CTA 磁吸 / 卡片 hover / 光标 / 点击反馈(如 Magnet / GlareHover / ClickSpark)
Component — 交互构件≥ 1卡片栈 / 画廊 / 导航 / 3D 卡片(如 CardSwap / ScrollStack / SpotlightCard / MagicBento)
Background — 氛围层≥ 1Hero 或全局背景(如 Aurora / Silk / Grainient / Threads)

L3 级别的累计 signature moments ≥ 6 个,但一页不超过 10 个(会躁)。

L3 电影级 scroll-story 硬性加码

L3 页面(用户说"像 doubao / apple / stripe 那样"、"要有 3D"、"普通"、"要电影感"等)除上述 6 类动效外,必须再至少覆盖 3/4 的 scroll-story 模式,详细模板见 references/scroll-story-patterns.md:

模式必备说明
Pin-Scrub 场景≥ 1section pin + 滚动驱动内容形变/时间线
左 pin / 右 swap 容器替换≥ 1左侧标题固定、右侧容器按滚动切换多场景(doubao 产品四连同款)
汇聚/散开转场≥ 1多元素从散落飞向合并 或 反之(doubao hero→次屏同款)
WebGL / 真 3D 签名时刻≥ 1Three.js / OGL / Babylon,不能只用 CSS 3D 充数

黄金判据:每 1-2 屏必有一个 signature moment。连续三屏只有 fade reveal 就不合格。

性能底线:

  • 单页 WebGL scene ≤ 1 个
  • 移动端自动降级(见 scroll-story-patterns.md 性能章节)
  • prefers-reduced-motion 必须有完整降级路径

风格适配:严格参照 references/motion-library.md 的「按风格 × 场景推荐组合」——Dark Editorial 不要用 GlitchText,Playful 不要用 Aurora,调性一致比酷炫重要。

性能底线:

  • 单页重背景(WebGL)≤ 1 个
  • 移动端自动降级:重背景 → 静态渐变;3D 组件 → 2D
  • Cursor 效果仅在 matchMedia('(hover: hover)') 下启用
  • 所有效果必须有 prefers-reduced-motion 降级路径

致谢:使用 vue-bits / reactbits 源码时,在页面 footer 或 README 注明 "Motion effects derived from vue-bits by DavidHDev (MIT)"。

参考物对齐审计(有参考时)
  • 已数清参考站 section 数量
  • 色板 / 字体 / 间距与参考数值对齐(或明确说明故意偏离的设计理由)
  • 参考站有滚动动效 → 必做 motion audit 并实现对应档位
  • 生成后显式做一次差异审计:列出哪些对齐、哪些故意不同

审计不是可选步骤。Phase C 代码写完必须逐条过一遍,未过的就地修。


技术栈原则

  • 框架:沿用项目已有栈;无则最小可行方案(静态 HTML 或 Vite)
  • 样式:沿用已有方案;不强行引入新依赖
  • 组件:有库复用;无库手写语义化组件
  • 动画:CSS 优先;L2 可用 IntersectionObserver;L3 允许 GSAP/ScrollTrigger/Lenis
  • 图片:禁止纯色块占位;占位图在代码顶部 const IMG 统一管理

语言规则

默认跟随输入语言,不强塞:

输入信号默认生成语言
参考 URL / 截图 / PRD 为中文中文
参考 URL / 截图 / PRD 为英文英文
用户对话语言作为兜底判定

多语言切换按需加:

  • 仅当用户明确要求,或 PRD 说明产品面向多市场时才加 i18n
  • 不要默认往每个页面塞 locale 对象和语言切换器——单语言产品里这是噪音
  • 需要 i18n 时:
    • 文案集中到独立 locale 对象(const messages = { zh: {...}, en: {...} }),组件内 t('hero.title') 引用
    • 切换即时生效,不要刷新
    • 简单项目:顶层 locale + React state;Next.js:next-intl / i18next;纯 HTML:<select> + data-lang

中文页面专属要求(生成中文内容时必守):

  • 中文字体用 Noto Sans SC / Noto Serif SC / LXGW WenKai 等中文字族,禁止只配英文字体让系统回退
  • 行高 ≥ 1.7,字距 letter-spacing: 0.02em
  • 正文字号 ≥ 15px,长篇阅读 ≥ 16px
  • 中英混排:中文字族在前,英文字族(Inter / DM Sans)作为 fallback,靠 font-family 链接入

参考文件索引

文件内容何时读取
references/design-md-template.mdDESIGN.md 输出模板(9 章节)Phase B
references/interaction-patterns.mdL1/L2/L3 三档交互代码库Phase A + B
references/motion-library.mdvue-bits/reactbits 效果目录 + 按风格推荐组合Phase B(选型)+ Phase C(落地)
references/scroll-story-patterns.mdL3 电影级 scroll-story 招式库(卡片星座 / 左pin右swap / 汇聚转场 / WebGL 签名)Phase B(L3 判定后)+ Phase C
references/style-seeds.md10 种风格种子 → Token 映射Phase A 对话时
references/scene-defaults.md7 种场景的布局/组件/交互基线Phase C
references/text-decoration-rules.md文字渐变/投影决策表Phase B
references/icon-library.mdlucide-react 图标速查Phase C
references/quality-checklist.md质量检查清单Phase C 审计
references/design-systems/INDEX.md58 个品牌设计系统索引Phase A 用户提到品牌时
references/design-systems/{name}.md单个品牌完整设计规范按需读取,禁止全量加载
scripts/crawl_website.pyPlaywright 爬虫(截图 + Token + 结构)Phase A 首选
scripts/extract_design_tokens.py轻量 Token 提取(静态站备选)Phase A 备选
scripts/fetch_unsplash_images.py占位图候选 URLPhase C

© xiaopu-ai, MIT. Rendered from Markdown: HTML in the file is shown as text, images as links, and headings moved down two levels. Raw file

Files

SKILL.md and 126 other files (scripts, references) in the repository root of xiaopu-ai/web-design.

  • SKILL.md
  • .gitignore
  • LICENSE
  • README.md
  • docs/DESIGN.md
  • docs/app.js
  • docs/images/optimized/01.webp
  • docs/images/optimized/02.webp
  • docs/images/optimized/03.webp
  • docs/images/optimized/04.webp
  • docs/images/optimized/05.webp
  • docs/images/optimized/06.webp
  • docs/images/optimized/07.webp
  • docs/images/optimized/08.webp
  • docs/images/optimized/09.webp
  • docs/images/optimized/10.webp
  • docs/images/optimized/11.webp
  • docs/images/optimized/12.webp
  • … and 109 more

Open the folder on GitHubat commit 22a4f48

Compare with similar skills

Web Design next to the 5 skills that share the most tags, products or categories with it. Stars are the repository's; “used in” counts other GitHub owners with a copy.

Web Design compared with similar skills
SkillStarsUsed inTokensAuto-checkLicenceRepo updated
Web Design this skillxiaopu-ai/web-design783—~3kAutomated safety check: PassMIT
Website Design Taste Extractorsenlindesign/taste-skill385—~5.3kAutomated safety check: PassNone
Extract Axure Datalintendo/Axhub-Skills130—~1.9kAutomated safety check: WarnMIT
Frontend Blueprinttech-leads-club/agent-skills7k1 repos~5.3kAutomated safety check: PassCC-BY-4.0
Product Skillsalirezarezvani/claude-skills28k—~2.6kAutomated safety check: PassMIT
Omk StitchKaimingWan/oh-my-kiro107—~2.2kAutomated safety check: PassMIT

Similar skills

  • Website Design Taste Extractor

    senlindesign/taste-skill

    Reverse-engineers a website's design from its URL into a Markdown and JSON pair holding concrete tokens and the trade-offs behind the look.

    385 GitHub stars~5.3k tokensUpdated 3 mo ago
    Frontend & DesignAuto-check passed
  • Extract Axure Data

    lintendo/Axhub-Skills

    Extract structured data from Axure prototypes using Playwright — screenshots, design tokens, interaction maps, annotations, and page text.

    130 GitHub stars~1.9k tokensUpdated 1 mo ago
    Frontend & DesignAuto-check: warnings
  • Frontend Blueprint

    tech-leads-club/agent-skills

    AI frontend specialist and design consultant that guides users through a structured discovery process before generating any code.

    7k GitHub starsUsed in 1 repo~5.3k tokens
    Frontend & DesignAuto-check passed
  • Product Skills

    alirezarezvani/claude-skills

    A skill your agent uses when coordinating product work across the 12 bundled product sub-skills (RICE, OKRs, UX research, design tokens, competitive teardown, analytics, experiments, discovery…

    28k GitHub stars~2.6k tokensUpdated 1 mo ago
    Frontend & DesignAuto-check passed
  • Omk Stitch

    KaimingWan/oh-my-kiro

    Google Stitch design-to-code workflow with design system management.

    107 GitHub stars~2.2k tokensUpdated 6 mo ago
    Frontend & DesignAuto-check passed
  • Design Dna

    zanwei/design-dna

    Extract, define, and apply design DNA across three dimensions: design system (tokens), design style (qualitative feel), and visual effects (Canvas, WebGL, 3D, particles, shaders, scroll effects…

    1.9k GitHub starsUsed in 1 repo~2.1k tokens
    Frontend & DesignAuto-check passed

Works with

Questions about Web Design

What does Web Design do?

Web 视觉设计 SKILL。输入 PRD / 参考 URL / 截图 / 关键词(任意组合),先产出一份标准化 DESIGN.md 设计规范,用户确认后据此生成 UI/UX、视觉、动效、响应式全部达标的 web 代码。专攻 web 端:Landing Page、Portfolio、产品页、博客、个人站、SaaS 介绍页等。当用户说"帮我做个网站""设计一个页面""参考 XX…. Web Design is an agent skill from xiaopu-ai/web-design.

When should I use Web Design?

Web Design fits situations like: tasks that involve PRD writing; tasks that involve Design tokens; tasks that involve Landing pages.

How do I install Web Design in Claude Code?

Run `npx skills add xiaopu-ai/web-design --skill web-design -a claude-code`. Or copy the skill folder (the xiaopu-ai/web-design repository) into .claude/skills/web-design in your project. Claude Code loads it when a task matches its description.

How do I install Web Design in Codex?

Run `npx skills add xiaopu-ai/web-design --skill web-design -a codex`. Or copy the skill folder (the xiaopu-ai/web-design repository) into .agents/skills/web-design in your project. Codex loads it when a task matches its description.

Can I use Web Design in Cursor, Gemini CLI or GitHub Copilot?

Cursor, Gemini CLI, GitHub Copilot and OpenCode also load SKILL.md folders. With the skills CLI, run `npx skills add xiaopu-ai/web-design --skill web-design -a cursor` (or -a gemini-cli, github-copilot or opencode for the others). To copy it by hand, put the folder in .cursor/skills/web-design, .gemini/skills/web-design, .github/skills/web-design and .opencode/skills/web-design in your project.

What does Web Design need to run?

Going by SKILL.md and its folder, Web Design needs JavaScript for the scripts in its folder and the command-line tools its instructions call (python3 and curl). Our summary lists: Python 3; Node.js.

Does Web Design access the network?

SKILL.md names 1 domain. As links in the text: github.com. This is read from the text; nothing was executed.

Is Web Design safe to install?

Our automated static check of SKILL.md found no risky patterns, such as piping downloads into a shell, reading credential files or hidden Unicode. It is not a guarantee. The check reads SKILL.md only: the scripts in the folder are not scanned, so read them before running anything.

What licence does Web Design use?

Web Design is published under the MIT licence (from the LICENSE file in the skill folder). It allows redistribution, so the full SKILL.md is shown on this page.

How many tokens does Web Design use?

About 3k tokens (SKILL.md is roughly 12k characters). Agents keep only the skill's name and description in context until a task matches; then they load SKILL.md in full. Its references folder adds about 246k tokens, read only when the agent opens those files.

What are the alternatives to Web Design?

Skills that share tags, products or a category with Web Design: Website Design Taste Extractor (senlindesign/taste-skill, 385 stars), Extract Axure Data (lintendo/Axhub-Skills, 130 stars), Frontend Blueprint (tech-leads-club/agent-skills, 7k stars) and Product Skills (alirezarezvani/claude-skills, 28k stars). The comparison table on this page puts their stars, adoption, token cost, safety result and licence side by side.

Who maintains Web Design?

xiaopu-ai (a GitHub user) maintains it in xiaopu-ai/web-design, which has 783 GitHub stars. The repository was last updated on June 24, 2026.

Source: xiaopu-ai/web-design on GitHub. Facts on this page come from the repository at the commit we read; the author's words are quoted as theirs.