wrysmil

Code Harness Kit — AI skill for Claude Code

AI community

Harness for Claude Code & Cursor · 让 AI 交付软件像一支工程团队——Leader 编排、阶段门禁、集体验证兜底.

How to install Code Harness Kit

This entry records only its repository, not the path inside it, so there is no exact command to give. Open wrysmil/Code-Harness-Kit and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What Code Harness Kit does

Harness for Claude Code & Cursor · 让 AI 交付软件像一支工程团队——Leader 编排、阶段门禁、集体验证兜底。Multi-agent orchestration with stage gates, 3-layer review, and Git Worktree isolation.

Alternatives in AI

  • Loushang — AI-native agent harness for coding workflows by python: multi-model LLM orchestration, stateful sessions, tool 1.3k ★
  • Pilotfish — Multi-model orchestration layer for Claude Code — the frontier model plans, cheaper models execute, verificati 688 ★
  • Claudexor — Multi-harness control plane for Claude Code, Codex, Cursor, and OpenCode: quota-aware rotation across multiple 425 ★

README

Code-Harness-Kit · 代码马具套件

![Code-Harness-Kit Banner — Production AI Agent Harness for Claude Code & Cursor](image/banner.png)

License: MIT Worker Agents Skills Platforms Language Install in 5 min

**Production-grade AI Agent Harness for Claude Code & Cursor** 把项目规则、路由判定、阶段门禁、过程产物、集体测试与三层并行审查、双层适配与工具链打包成一套脚手架 —— 让 LLM 像一支有纪律的工程团队一样交付软件,而不是"单 Agent 一把梭哈"。

**核心价值**:解决 LLM 代码生成"可靠性、一致性、长期可维护性"三大顽疾。Harness 像马具 —— 用缰绳与鞍具把马力引到正确方向,而不是限制能力本身。

**核心理念**:**Agent Team · 阶段门禁 · 集体测试 · 集体审查 · 文档驱动**——五件套协同。


TL;DR

  • 🎯 给 AI 上缰绳:把"AI 单兵写代码"升级为"工程团队协作",自带阶段门禁与尾盘强制审查
  • 🤖 跨平台适配:同一套规则同时跑在 Claude Code / Cursor / Trae,业务项目零侵入接入
  • 🌳 Git Worktree 隔离:每个 WU 独立 worktree,文件所有权清晰,并行不打架
  • 🧪 集体测试 + 三层审查:批次收尾强制 code review / OWASP / CWV 并行扇出,不靠"应该通过"
  • 📜 文档驱动:所有过程产物落盘 .ai-runtime-artifacts/,跨会话可接续

**30 秒判断要不要用**:你在用 Cursor / Claude Code 写代码,但担心 LLM 跑偏、PR 质量不稳 → 装上它。如果你只是写单文件 typo,它太重 —— 用 Claude Code 原生即可。


目录