Auto-Coding Skill

Version: 1.0.0
Author: Krislu + Claw Soft
Description: Multi-Agent Autonomous Coding System

---

📖 Overview / 概述

Auto-Coding is a multi-agent collaboration system built on OpenClaw that automatically transforms user requirements into runnable applications.

Auto-Coding 是一个基于 OpenClaw 的多 Agent 协作系统，可以自动将用户需求转化为可运行的应用程序。

---

🎯 Features / 功能特性

• Multi-Agent Collaboration - Initializer + Coder dual-agent pattern 多 Agent 协作 - 初始化器 + 编码器双代理模式

• Automatic Task Breakdown - Split requirements into 20-50 features 自动任务分解 - 将需求拆分为 20-50 个功能点

• Progress Tracking - Real-time status updates with Git integration 进度追踪 - 实时更新状态，集成 Git 版本控制

• Security Sandbox - Command allowlist with 40+ safe commands 安全沙箱 - 40+ 安全命令白名单机制

---

🚀 Usage / 使用方法

Command Triggers / 命令触发

# English commands
/auto-coding create a Todo app
/auto-coding build a weather app

# Chinese commands
/auto-coding 创建一个 Todo 应用
/auto-coding 帮我开发一个天气应用

# Natural language
帮我创建一个 博客系统
Help me create a blog system

Python API / Python 接口

from auto_coding import AutonomousCodingController

controller = AutonomousCodingController(
    project_name="my-app",
    requirements="Create a Todo application"
)

result = await controller.run_full_cycle(max_iterations=10)
print(f"Completed: {result['completed']}/{result['total']} tasks")

---

📁 Project Structure / 项目结构

auto-coding/
├── skill.py                    # OpenClaw Skill implementation
├── agent_controller.py         # Main controller
├── task_manager.py             # Task management
├── security.py                 # Security configuration
├── test_auto_coding.py         # Test scripts
├── prompts/
│   ├── initializer.md          # Initializer agent prompt
│   └── coder.md                # Coder agent prompt
└── README.md                   # Documentation

---

🛡️ Security / 安全性

Command Allowlist / 命令白名单

Reference Implementation / 参考实现:

• security.py defines recommended command allowlists (40+ safe commands)
• security.py 定义了推荐的命令白名单 (40+ 安全命令)

Note / 注意:

• This is a reference implementation for documentation purposes
• 这是用于文档目的的参考实现
• Actual security enforcement depends on OpenClaw Gateway runtime
• 实际安全执行依赖 OpenClaw Gateway 运行时

Allowed Commands / 允许的命令 (40+):

Category / 类别	Commands / 命令
File Inspection / 文件检查	ls, cat, head, tail, grep, find
File Operations / 文件操作	cp, mkdir, chmod, touch, mv
Node.js	npm, npx, node, yarn, pnpm
Python	pip, python, pytest, unittest
Version Control / 版本控制	git
Process Management / 进程管理	ps, sleep, pkill (dev processes only)

Blocked Commands / 禁止的命令

• ❌ sudo - Privilege escalation
• ❌ rm -rf - Forced recursive deletion
• ❌ dd, mkfs - Disk operations
• ❌ systemctl, service - System services

---

🧪 Testing / 测试

cd auto-coding/

# Run all tests
python3 test_auto_coding.py 1  # Minimal test
python3 test_auto_coding.py 2  # Real project test
python3 test_auto_coding.py 3  # All tests

Expected Output / 预期输出

🧪 Auto-Coding Test
==================================================
✅ TaskManager tests passed
✅ Security tests passed
✅ Controller tests passed
🎉 All tests passed!

---

📋 Configuration / 配置

Environment Variables / 环境变量

# Project root directory (optional)
export AUTO_CODING_PROJECTS_DIR="/path/to/projects"

# Max iterations (optional)
export MAX_ITERATIONS="50"

---

⚠️ Known Limitations / 已知限制

1. OpenClaw Integration - Requires sessions_spawn support from OpenClaw Gateway OpenClaw 集成 - 需要 OpenClaw Gateway 的 sessions_spawn 支持

2. Mock Mode - Currently runs in mock mode without real Gateway configuration 模拟模式 - 当前在没有真实 Gateway 配置的情况下运行在模拟模式

3. Language Support - Best results with Chinese and English prompts 语言支持 - 中文和英文提示效果最佳

---

📚 Documentation / 文档

• README.md - Full user guide with examples
• prompts/initializer.md - Initializer agent instructions
• prompts/coder.md - Coder agent instructions
• Security-Compliance-Report.md - Security audit report

---

📄 License / 许可证

Internal Use Only - Lobster Team

---

👥 Authors / 作者

• Krislu - Project Initiator
• Claw Soft - Lead Developer

---

Last Updated: 2026-03-14