---

name: edgeone-pages-website-skeleton description: >- 基于 EdgeOne Pages 的全栈网站生成方案。用户说一句话（如「帮我建一个电商站」「做一个AI客服站」「做个管理后台」），AI 自动组合 Auth、Cart、Payment、AI Chat、Admin 五大模块，生成完整 Next.js 前后端代码并部署到 EdgeOne Pages 全球 CDN。支持电商、AI 助手、SaaS 管理后台三大模板。底层使用 Edge Functions（KV 存储、JWT 校验）+ Cloud Functions（bcrypt、MySQL、微信/支付宝支付、SSE 流式 AI）+ Platform Middleware（鉴权、CORS、CSP、支付回调 IP 白名单）。支持支付幂等原子锁、订单超卖防护、Refresh Token 并发安全等 P0 安全设计。

触发场景： "帮我建一个电商网站"、"帮我建一个AI客服站"、"做个带登录的电商站"、"建一个全栈网站"、 "帮我做一个带支付功能的网站"、"做一个电商站，包含登录注册、购物车、微信支付宝支付"、 "Build me an e-commerce website"、"Create an AI customer service site with login"、 "Make a SaaS admin dashboard"、"建一个网站，登录注册、AI客服、支付全要" metadata: author: liuboacean version: "2.1.0" demo: https://geek-mall-demo-4qaxvmeh.edgeone.cool

Website Skeleton Skill — EdgeOne Pages 全栈网站骨架

Scaffold and deploy a production-ready full-stack website on EdgeOne Pages from a single natural language request.

When to use this skill

• User wants to start a new full-stack website (e-commerce, AI chatbot, SaaS admin) with zero config
• User mentions: e-commerce, shopping cart, payment, AI chat, admin dashboard, login/register, or similar
• User wants Edge Functions + Cloud Functions + KV + MySQL on EdgeOne Pages

Do NOT use for:

• Adding a single API/function to an existing project → use edgeone-pages-dev
• Deploying an already-built project → use edgeone-pages-deploy
• Pure static sites or blogs without business logic

---

⛔ Critical Rules (never skip)

1. ALWAYS ask the user to confirm the scenario first. Three built-in templates are available:
   • E-commerce (🛒)：Auth + Cart + Payment (WeChat/Alipay) + Orders + Admin
   • AI Assistant (🤖)：Auth + AI Chat (SSE streaming) + Admin
   • SaaS Admin (📊)：Auth + RBAC Admin + Subscription Payments
2. Never fabricate the user's requirements. If information is missing, ask clarifying questions before generating code.
3. Respect EdgeOne Pages platform constraints:
   • KV is only accessible from Edge Functions (not Cloud Functions)
   • Cloud Functions directory must be named cloud-functions/
   • bcrypt must run inside Cloud Functions
   • AI SSE streaming must be implemented in Cloud Functions (Edge cannot use waitUntil)
4. Payment idempotency is mandatory. Use Edge putIfNotExists with 24h TTL (< WeChat retry window of 72h).
5. Never commit .env.local or any real API keys. Ensure .gitignore covers it.
6. Always verify EdgeOne Pages CLI is installed before deployment.

---

Main Flow

Step 1 — Confirm scenario

Use ask_followup_question to ask:

你想要建什么类型的网站？

• 🛒 电商站 — 登录注册、购物车、微信/支付宝支付、订单管理
• 🤖 AI 客服站 — 登录注册、AI 对话（SSE 流式）、访客留言
• 📊 SaaS 管理后台 — 登录注册、RBAC 权限、数据看板
• ⚙️ 自定义 — 自由组合模块

Branch:

• E-commerce → use references/templates.md § "电商模板"
• AI Assistant → use references/templates.md § "AI助手模板"
• SaaS Admin → use references/templates.md § "管理后台模板"
• Custom → confirm which modules are needed

Step 2 — Collect requirements

Read references/templates.md and confirm with user:

1. Site name / domain
2. Which modules to enable (Auth always required)
3. Payment providers (WeChat / Alipay / Both / None)
4. AI provider (if AI Chat enabled)

Output: a Spec object

{
  "scenario": "e-commerce | ai-assistant | saas-admin | custom",
  "siteName": "...",
  "modules": ["auth", "cart", "payment", "ai-chat", "admin"],
  "paymentProviders": ["wx", "ali"],
  "aiProvider": "openai | hunyuan | none",
  "envVars": ["JWT_SECRET", "DATABASE_URL", ...]
}

Show the Spec back and ask for confirmation.

Step 3 — Generate project structure

Based on the selected template, generate the following directory tree:

<project>/
├── client/                     # Next.js frontend (SPA)
│   ├── app/ (pages: login, register, cart, checkout, orders, admin/*)
│   ├── lib/ (api client, auth service, cart service, SSE client)
│   └── middleware (optional, for pure SPA)
├── edge-functions/            # Edge Functions (V8 + KV)
│   ├── _middleware.js         # JWT check + KV session + rate limit
│   └── api/
│       ├── auth/login.js     # JWT issue + KV session
│       ├── auth/me.js        # KV session read
│       ├── auth/refresh.js   # RT rotation with KV version lock
│       ├── auth/logout.js
│       ├── internal/idempotency.js  # Edge atomic idempotency lock ← P0
│       ├── products/list.js  # KV cache + MySQL fallback
│       ├── cart/*.js         # KV cart
│       ├── orders/list.js    # MySQL read
│       └── ai/history.js     # KV AI session history
├── cloud-functions/          # Cloud Functions (Node.js + MySQL)
│   ├── api/
│   │   ├── auth/register.js  # bcrypt cost=12 ← P0
│   │   ├── pay/create-order.js
│   │   ├── pay/wx-notify.js  # Edge idempotency lock ← P0
│   │   ├── pay/ali-notify.js
│   │   ├── order/create.js   # SELECT FOR UPDATE ← P0
│   │   ├── order/cancel.js
│   │   ├── admin/products.js
│   │   └── ai/chat-stream.js  # SSE streaming ← AI module
│   └── utils/
│       ├── db.js             # MySQL pool (mysql2/promise)
│       └── payment-sdk.js    # WeChat V3 / Alipay SDK
├── db/
│   ├── migrations/001_init.sql  # users, products, orders, order_items
│   └── seed.sql
├── docs/
│   └── env-vars.md           # Environment variable matrix
└── references/
    └── auth-module.md / payment-module.md / ai-chat-module.md / kv-storage.md

Generate each file according to references/ documents.

Step 4 — Environment variables

Read references/env-setup.md (or inline from SKILL.md § Environment Variables) and:

1. List all required env vars for the selected modules
2. Ask user to provide values (or provide placeholder examples)
3. Remind: never commit .env.local

Step 5 — Local verification

Guide user to run locally:

npm run dev   # Next.js on http://localhost:3000

Verify:

• Homepage renders
• Auth signup/login works
• Selected modules (cart / payment / AI chat) are functional

Step 6 — Deploy to EdgeOne Pages

6.1 — Check edgeone-pages-deploy skill

Check if ~/.codebuddy/skills/edgeone-pages-deploy/ exists (or Windows equivalent). If yes → go to 6.2. If no → go to 6.3.

6.2 — Hand off to deploy skill

✅ 项目已就绪，本地跑通后随时可以部署。

告诉 AI 「部署到 EdgeOne Pages」，会自动加载 edgeone-pages-deploy skill 完成上线。

6.3 — Manual deploy

# Install CLI (first time)
npm install -g edgeone@latest

# Login
edgeone login --site china

# Deploy
cd <project>
edgeone pages deploy -n <project-name>

⚠️ The final URL contains eo_token / eo_time query params — must be kept, removing them causes 401.

---

Routing

Step	Read
Scenario confirmation + module matrix	references/templates.md
Auth module (JWT, KV session, bcrypt, RT)	references/auth-module.md
Payment module (idempotency, WeChat/Alipay, callbacks)	references/payment-module.md
AI Chat module (SSE, history, streaming)	references/ai-chat-module.md
KV Storage strategy (single query, layered access)	references/kv-storage.md

---

Quick Reference

Demo: https://geek-mall-demo-4qaxvmeh.edgeone.cool

Platform constraints:

• KV → Edge Functions only
• bcrypt → Cloud Functions only
• AI SSE → Cloud Functions only
• Platform Middleware → CORS, CSP, Bearer check, Payment callback IP whitelist (direct return)

P0 Security (all implemented):

• Payment idempotency: Edge putIfNotExists (24h TTL)
• Order atomicity: SELECT FOR UPDATE + optimistic lock + MySQL CHECK
• RT concurrency: KV version optimistic lock (409 retry)
• Price integrity: Server-side MySQL (never trust frontend price)
• Payment callback isolation: Platform Middleware direct return
• Password: bcrypt cost ≥ 12

Template repos: https://github.com/TencentEdgeOne/edgeone-pages-skills

---

⚠️ 安全与合规说明

1. 部署确认机制

AI Agent 部署前会显示待部署文件清单和变更概要，请求用户确认后执行部署。

2. 支付安全

• 支付回调实现 HMAC-SHA256 签名验证（微信官方算法）
• 校验 appid/mchid 商户身份一致性
• 校验回调金额与订单金额完全匹配
• 仅接受 result_code=SUCCESS 的交易状态
• KV 幂等锁防止重复回调处理

3. 数据保留策略

数据类型	存储位置	保留期限
KV Session	EdgeOne KV	7 天 TTL
AI 聊天历史	EdgeOne KV	30 天 TTL
审计日志	EdgeOne KV	90 天 TTL
订单数据	MySQL	永久（业务必需）

4. Cron 定时任务

• PENDING 超时 30 分钟 → CANCELLED
• SHIPPED 超过 7 天 → COMPLETED
• 提供一键禁用脚本 cloud-functions/cron/disable-cron.js

5. 使用建议

• 在测试项目中验证全部功能
• 使用测试/沙箱商户号
• 部署前审查生成的代码和环境变量
• 启用真实支付前完成签名验证测试