一套面向 0→1 产品的 Claude Code Skills 套件:从市场需求到设计规范的全链路文档生成器,最终产出可直接喂给 Claude Code / Cursor 等 AI 编程 Agent 执行的项目规范。
A suite of Claude Code Skills for 0→1 product development: an end-to-end documentation chain that turns a raw idea into specs ready to feed into Claude Code / Cursor.
| Skill | 中文说明 | English |
|---|---|---|
mrd-writing |
市场需求文档生成器 — 从真实用户数据中提炼市场信号 | Market Requirement Doc — extract market signals from real user data |
brd-writing |
商业需求文档生成器 — 帮你在投入时间前判断方向值不值得做 | Business Requirement Doc — decide if a direction is worth pursuing before you sink time into it |
prd-writing |
项目规范生成器 — 把想法转化为 10 模块完整 PRD,可直接喂 AI 编程 Agent | Project spec generator — turn an idea into a 10-module PRD ready for AI coding agents |
design-spec |
设计规范生成器 — 读 PRD 产出包含色板/字体/间距/动效的 DESIGN.md | Design spec generator — read the PRD and produce a DESIGN.md with full design tokens |
这四个 skill 不是独立工具,而是一条有先后顺序的决策链。每一步的输出是下一步的输入。
These four skills are not standalone tools — they form a sequential decision chain, where each step's output feeds the next.
/mrd 分析市场需求 → MRD.md
analyze market needs
↓
/brd 判断商业可行性 → BRD.md (reads MRD.md)
assess feasibility
↓
/prd-writing 定义项目结构 → PRD/ (reads BRD.md + MRD.md)
define product spec
↓
/design-spec 产出视觉规范 → DESIGN.md (reads PRD/)
produce design tokens
↓
Claude Code 落地实现 MVP 代码 (reads PRD/ + DESIGN.md)
build the MVP
-
🧬 自动继承,不重复提问 — 下游 skill 自动读取上游的「交接区」,已确认的信息不再问第二遍。 Auto-inheritance, no re-asking — downstream skills auto-read the upstream "handoff section"; confirmed info is never re-asked.
-
🚦 健康度闸门 — 每一步都有质量等级标注。上游证据弱(🔴) → 下游最高只能(🟡),防止跑偏。 Health gates — every step has an evidence-level tag. Weak upstream (🔴) caps downstream at (🟡), preventing drift.
-
🚧 PRD 与 DESIGN 的硬性边界 — PRD 决定结构(页面/组件/路由/数据),DESIGN 决定视觉(颜色/字体/间距/动效)。两层分离,互不污染。 Hard PRD↔DESIGN boundary — PRD owns structure (pages/components/routes/data); DESIGN owns visuals (colors/fonts/spacing/motion). Strict separation, no cross-contamination.
-
📦 默认 Web 端 — 所有产出默认面向 Web 产品(移动端优先的响应式网页),除非用户明确说做 App。 Web-first by default — outputs target Web products (mobile-first responsive) unless the user explicitly asks for a native app.
git clone https://github.com/limengzhe27-boop/claude-product-doc-skills.git
mkdir -p ~/.claude/skills
cp -R claude-product-doc-skills/skills/* ~/.claude/skills/在你的项目根目录: In your project root:
git clone https://github.com/limengzhe27-boop/claude-product-doc-skills.git
mkdir -p .claude/skills
cp -R claude-product-doc-skills/skills/* .claude/skills/安装后重启 Claude Code,在对话中用 /mrd、/brd、/prd-writing、/design-spec 触发,或用自然语言描述需求(例如"帮我写个 PRD")自动激活。
After install, restart Claude Code. Trigger with /mrd, /brd, /prd-writing, /design-spec, or just describe your intent in natural language (e.g. "help me write a PRD") to auto-activate.
# 从 0 开始一个新产品 / Starting a new product from scratch
You: /mrd # 喂用户访谈数据,产出 MRD.md
# feed user interviews → MRD.md
You: /brd # 判断方向是否值得做,产出 BRD.md
# decide if it's worth doing → BRD.md
You: /prd-writing # 定义完整项目规范,产出 PRD/ 文件夹
# define full project spec → PRD/
You: /design-spec # 基于 PRD 产出视觉规范,产出 DESIGN.md
# produce design spec → DESIGN.md
You: 帮我实现 MVP / Build the MVP # Claude Code 读 PRD/ + DESIGN.md 写代码
# Claude Code reads PRD/ + DESIGN.md → code
你也可以从任意一步开始——例如已经有清晰的方向,直接 /prd-writing 即可,skill 会自适应。
You can also start from any step — e.g. if you already have a clear direction, jump straight to /prd-writing and the skill will adapt.
claude-product-doc-skills/
├── README.md
└── skills/
├── mrd-writing/
│ └── SKILL.md
├── brd-writing/
│ └── SKILL.md
├── prd-writing/
│ ├── SKILL.md
│ └── references/
│ ├── module-specs.md # PRD 10 模块规范细节
│ └── style-guide.md # PRD 风格指南
└── design-spec/
└── SKILL.md
✅ 适合 / Good for
- 0→1 新产品立项,需要把模糊想法变成可执行规范
- 个人开发者 / 独立产品人想用 AI 编程 Agent vibe-code 一个 MVP
- 需要在投入开发前先做一轮严肃的可行性评估
- 0→1 new products turning fuzzy ideas into executable specs
- Solo devs / indie makers vibe-coding MVPs with AI agents
- Anyone who wants a serious feasibility check before writing code
⛔ 不适合 / Not for
- 已有成熟产品的功能迭代(用更轻量的工单流程即可)
- 纯技术重构 / Bug 修复
- B 端复杂业务系统的正式 PRD(本套面向轻量 Web 产品)
- Feature iteration on mature products (use a lighter ticket flow)
- Pure refactors / bug fixes
- Heavy enterprise B2B PRDs (this suite targets lightweight web products)
MIT
发现问题或有改进想法?欢迎提 Issue 或 PR。 Issues and PRs welcome.