这是一个跨 Agent 使用的工程技能包，用于审查、调试、开发和发布基于腾讯云 CloudBase 的原生微信小程序。核心规范采用 SKILL.md，平台差异集中在独立适配文档中。
- 分层检查 WXML、WXSS、JavaScript、JSON、页面状态、云函数、数据库约束和部署配置。
- 定位超时、权限、支付、订单、退款、物流、定时对账和云控制台配置问题。
- 使用无第三方依赖、只读运行的静态扫描器发现常见结构缺陷和高风险模式。
- 区分已经由代码和日志证明的问题，以及只能通过微信开发者工具、真机或云控制台验证的事项。
兼容性由加载 Skill 的运行环境或客户端决定，不由底层模型名称单独决定。
| 环境 | 使用方式 | 支持级别 |
|---|---|---|
| Codex | 读取同一份 SKILL.md |
原生 |
| Claude Code | 安装到 Claude Skills 目录 | 原生 |
| Claude.ai | 在支持自定义 Skill 的账户中上传 ZIP | 产品功能 |
| Kimi Code CLI | 从通用或 Kimi 技能目录发现 | 原生 |
| DeepSeek API | 由支持 Skill 的 host 或自建运行时加载 | 间接 |
| 普通网页聊天 | 提供通用提示词和必要项目文件 | 手动回退 |
具体安装方法：
不能访问文件系统、终端或云控制台的网页聊天只能分析用户提供的材料，不能真正扫描、修改、运行测试或部署项目。
需要 Node.js 18 或更高版本。
npm test
node scripts/scan-miniprogram.mjs /path/to/project
node scripts/scan-miniprogram.mjs /path/to/project --json扫描结果是人工复核线索，不能替代微信开发者工具编译、真机测试、云函数日志和受控线上验证。
技能要求遵循目标仓库的现有约定，并区分代码缺陷与平台配置问题。本仓库不包含任何特定应用的源码、标识符、凭据、业务数据或部署密钥。
官方兼容性资料：OpenAI Skills、Claude Agent Skills、Kimi Code Skills、DeepSeek 编程代理接入。
A reusable engineering skill for auditing, debugging, implementing, and releasing native WeChat Mini Programs backed by Tencent CloudBase. The tool-neutral workflow lives in SKILL.md; host-specific installation guidance lives in separate adapters.
- Layered review of WXML, WXSS, JavaScript, JSON, client state, cloud functions, database invariants, and deployment configuration.
- Debugging guidance for timeouts, permissions, payments, orders, refunds, shipping, scheduled reconciliation, and cloud-console configuration.
- A dependency-free, read-only static scanner for common structural defects and risky patterns.
- Explicit separation between locally proven findings and checks that require WeChat developer tools, a real device, or a live cloud console.
Compatibility is determined by the host or client that loads the skill, not by the model name alone.
| Host | Method | Support |
|---|---|---|
| Codex | Load the same SKILL.md |
Native |
| Claude Code | Install in a Claude Skills directory | Native |
| Claude.ai | Upload a ZIP where custom Skills are available | Product feature |
| Kimi Code CLI | Discover from a generic or Kimi skill directory | Native |
| DeepSeek API | Use a skill-aware host or custom runtime | Indirect |
| Plain web chat | Supply the fallback prompt and relevant files | Manual fallback |
Installation guides: Codex, Claude, Kimi Code, DeepSeek and other models, and the generic system prompt.
A web chat without filesystem, shell, or cloud-console tools can analyze supplied material, but it cannot genuinely scan, edit, test, or deploy a repository.
Node.js 18 or newer is required.
npm test
node scripts/scan-miniprogram.mjs /path/to/project
node scripts/scan-miniprogram.mjs /path/to/project --jsonScanner findings are leads for manual review. They do not replace WeChat developer-tool compilation, real-device testing, cloud logs, or controlled live validation.
The skill preserves repository conventions and separates code defects from platform configuration. This repository contains no application-specific source, identifiers, credentials, business data, or deployment secrets.