Coding

Multi Agent Codegen

Try it

用 4 Agent 半串行协作流水线(Plan/Write/Test/Refine)把老板的一句话需求变成可用的 Python 软件。Use when user says "做个软件", "开发工具", "写个脚本", or wants software built by AI agents.

What it does

用 4 Agent 半串行协作流水线(Plan/Write/Test/Refine)把老板的一句话需求变成可用的 Python 软件。Use when user says "做个软件", "开发工具", "写个脚本", or wants software built by AI agents.

The skill document

Multi-Agent Codegen 🐛

4 Agent 软件工程协作流水线(按老板 2026-06-22 定义的标准)

把老板的一句话需求,经过 Plan → Write → Test → Refine 四个 agent 协作,自动生成完整可运行的 Python 软件 + 测试用例 + 质量评审。

4 Agent 角色(老板标准定义)

Agent角色输入输出
🏛️ Plan架构师需求PRD(单文件实现的设计)
💻 Write开发者PRD + Refine 反馈完整 Python 代码
🧪 Test测试工程师PRD + Codepytest 测试(看到完整代码)
🔍 Refine审查者Code + Tests评分 0-100 + 改进意见

架构(半串行 + 循环)

START → Plan → Write → Test → Refine → 条件分支
                                       ├─ score >= 70 → END
                                       └─ score < 70 → Write(带 Refine 反馈重写)

v2.1 验证数据(基于 todo 工具需求):

  • 测试通过率:39/40 (97.5%)
  • 评分:92/100(一次过)
  • 耗时:~3 分钟

快速使用

# 全局命令(任意目录)
multi-codegen "做个命令行 todo 工具"
multi-codegen "做个图片压缩脚本"
multi-codegen "做个 Web 爬虫"

# JSON 模式(给 agent / cron 调用)
IDEA_GEN_JSON=1 multi-codegen "做个 API 客户端"

输出产物

保存到 ~/.openclaw/workspace-coding-advisor/output/multi_agent_codegen/

plan.md        # Plan 架构设计
code.py        # Write 完整代码
test_code.py   # Test pytest 用例
refine.md      # Refine 评审意见

技术栈

组件版本作用
LangGraph1.2.6StateGraph 状态机
langchain-anthropic1.4.6走 Anthropic 协议调 M3
M3 (MiniMax-M3)-所有 4 个 agent 共用
Python3.10+-

依赖

首次运行会自动检测并安装:

  • langgraph >= 1.2.0
  • langchain-anthropic >= 1.4.0
  • pytest >= 9.0.0(验证用)

如需手动安装:pip install -r requirements.txt

与其他 skill 配合

  • langgraph-idea-generator(快速出 3 行方案)→ multi-agent-codegen(完整软件)
  • idea-gen 评估可行性,再 multi-codegen 真正开发

文件结构

multi-agent-codegen/
├── SKILL.md
├── skill.json
├── _meta.json
├── .venv → ../../.venv-langgraph(共享 venv)
├── scripts/
│   ├── multi_codegen.py    # 完整 4 agent 流水线
│   └── cli.sh              # CLI 入口
└── .clawhub/origin.json

关键约束(防 LLM bug)

  • 单文件实现:所有代码一个 .py 文件
  • 方法名避免 builtin:list/dict/str/int/type → 用 list_tasks/get_items
  • retry 完整文件:Write 改进时必须输出完整新版本
  • 测试看完整代码:Test 在 Write 之后启动(半串行,不是真并行)

适用 vs 不适用

✅ 老板说"做个软件/工具/脚本" ✅ 复杂需求需要 PRD + 代码 + 测试 + 审查 ❌ 1-2 行小脚本(单 agent 更快) ❌ 纯分析/报告任务(非软件工程)

相关 Rule

  • [RULE-20260622-001] 4-Agent 软件工程协作流水线标准工作流(MEMORY.md)

版本: v1.0.0 作者: 码虫 coding-advisor 创建日: 2026-06-22

Related skills

Govern any task that uses or considers multiple agents, subagents, delegation, parallel work, or speed-first execution. Use whenever the user says "多派 Agent"...

1 installs

Coordinate multi-agent swarms for parallel and pipeline workflows. Use when coordinating multiple agents, running parallel reviews, building pipeline workflows, or implementing divide-and-conquer patterns with subagents.

27 installs

Design evaluation suites, interpret agent benchmarks, and compare CrewAI, LangChain, AutoGen, and others.

22 installs1 stars

Use when an AI coding agent needs bounded development loops, persistent project-local Docs/ state, context budgeting, environment escalation rules, safe stop...

by EnglandTong

Create or refactor production-grade TypeScript agent applications. Use when the user asks Codex to generate, scaffold, restructure, or harden a TypeScript agent app, including CLI agents, web agent apps, API services, internal tools, multi-agent harnesses, workflow-first systems, model provider wiri

A comprehensive governance system for managing agent roles, responsibilities, handoff policies, and conflict resolution in multi-agent environments