Create Harness Docs

智能分析项目结构，自动创建符合 Harness Engineering 要求的文档体系。支持 Spring Boot、React、Vue、NestJS、Express、Django、FastAPI、Go 等多种项目类型。

Security Scan

The skill's code and instructions match its stated purpose (analyzing a repo and generating Harness-style docs); it runs local filesystem operations and shell commands but requests no external credentials or installs.

Purpose & Capability

The name/description promise to detect project types and create documentation; the included script implements project-type detection (package.json, pom.xml, go.mod, manage.py, etc.), scans source folders, and generates docs/ and CI/workflow files. Required resources (none) align with this purpose.

Instruction Scope

SKILL.md instructs running the bundled node script with flags like --init/--agents/--validate. The script performs local operations: reading project files, scanning directories, running find via execSync, creating directories and writing documentation and CI/test files. This stays within the claimed scope, but be aware it will modify the repository (create docs/, .github/workflows/, tests, etc.).

Install Mechanism

There is no install spec; this is instruction-only plus one bundled Node script. Nothing is downloaded or executed from remote URLs. No package installs or external installers are invoked by the skill itself.

Credentials

The skill declares no environment variables, no credentials, and does not attempt to read config paths outside the project. It does not request secrets or unrelated external access — its filesystem access is proportional to its purpose.

Persistence & Privilege

always:false (default) and normal autonomous invocation settings. The skill does write files into the current project (docs, workflows, tests) which is expected behavior; it does not request persistent system-wide privileges or modify other skills' configs.

Assessment

This skill appears internally consistent, but it will write files and run shell commands inside the repository you run it in. Before running: (1) review the script (scripts/create-harness-docs.js) yourself to confirm you accept the exact changes; (2) run it first in a copy/branch or test repo (or use the smaller flags like --agents) so you can inspect generated files; (3) back up or commit your repo before running; (4) note there are execSync/find calls and filesystem writes — if you need absolute assurance of no network activity, inspect the full script for fetch/HTTP calls (none were detected in the provided portion). If you want extra safety, run the script inside a disposable container or CI job so changes are contained.

Create Harness Docs (智能版)

自动分析当前项目类型并创建对应的 Harness Engineering 文档体系。

功能

1. 智能检测 - 自动识别项目类型 (Spring Boot/React/Vue/NestJS 等)
2. 生成对应模板 - 根据项目类型生成相应的分层架构、CI 配置
3. 创建文档 - AGENTS.md、架构文档、质量评级等
4. 架构测试 - 为 Java 项目生成 ArchUnit 测试

支持的项目类型

类型	检测依据	分层架构
Spring Boot	pom.xml	Entity→DAO→Service→Controller
React	package.json + react	Types→Components→Hooks→Pages
Vue	package.json + vue	Types→Components→Composables→Views
NestJS	@nestjs/core	Entities→Repositories→Services→Controllers
Express	express	Types→DAO→Service→Controller
Django	manage.py	Models→Views→Serializers→Urls
FastAPI	fastapi	Models→Schemas→Services→Routes
Gin	go.mod	Models→Repository→Service→Handler

使用方法

# 在项目目录运行，自动检测类型并创建文档
node create-harness-docs.js --init

# 选项
# --init          创建所有文档
# --agents        仅创建 AGENTS.md
# --architecture  仅创建架构文档
# --quality       仅创建质量评级
# --validate      验证现有文档

创建的文件

项目目录/
├── AGENTS.md                     # 入口/索引
├── docs/
│   ├── architecture/
│   │   ├── ARCHITECTURE.md      # 架构总览
│   │   └── domains/             # 业务域详情
│   ├── plans/                   # 执行计划
│   └── quality/grades.md        # 质量评级
├── .github/workflows/
│   └── harness-ci.yml          # CI 配置
└── (Java) src/test/.../ArchitectureTest.java  # ArchUnit 测试

核心原则

遵循 OpenAI Harness Engineering:

1. AGENTS.md 是目录，不是手册 - 知识放 docs/
2. 严格分层 - 根据项目类型定义的分层架构
3. 约束即代码 - 用 linter/ArchUnit 强制规则
4. 持续清理 - 定期处理技术债务