#!/usr/bin/env python3 """ Agent 脚手架脚本 - 按最佳实践创建新的 Agent 项目。 用法： python init_agent.py [--level 0-4] [--path ] 示例： python init_agent.py my-agent # Level 1（4 个工具） python init_agent.py my-agent --level 0 # 最小版（仅 bash） python init_agent.py my-agent --level 2 # 带 TodoWrite python init_agent.py my-agent --path ./bots # 自定义输出目录 """ import argparse import sys from pathlib import Path # 各等级的 Agent 模板 TEMPLATES = { 0: '''#!/usr/bin/env python3 """ Level 0 Agent - Bash 就够了（约 50 行） 核心洞察：一个工具（bash）就能做很多事。 通过自递归启动子 Agent：python {name}.py "subtask" """ from anthropic import Anthropic from dotenv import load_dotenv import subprocess import os load_dotenv() client = Anthropic( api_key=os.getenv("ANTHROPIC_API_KEY"), base_url=os.getenv("ANTHROPIC_BASE_URL") ) MODEL = os.getenv("MODEL_NAME", "claude-sonnet-4-20250514") SYSTEM = """你是一个编码 Agent。所有事情都通过 bash 完成： - 读取：cat, grep, find, ls - 写入：echo 'content' > file - 子 Agent：python {name}.py "subtask" """ TOOL = [{{ "name": "bash", "description": "执行 shell 命令。", "input_schema": {{"type": "object", "properties": {{"command": {{"type": "string"}}}}, "required": ["command"]}} }}] def run(prompt, history=[]): history.append({{"role": "user", "content": prompt}}) while True: r = client.messages.create(model=MODEL, system=SYSTEM, messages=history, tools=TOOL, max_tokens=8000) history.append({{"role": "assistant", "content": r.content}}) if r.stop_reason != "tool_use": return "".join(b.text for b in r.content if hasattr(b, "text")) results = [] for b in r.content: if b.type == "tool_use": print(f"> {{b.input['command']}}") try: out = subprocess.run(b.input["command"], shell=True, capture_output=True, text=True, timeout=60) output = (out.stdout + out.stderr).strip() or "（无输出）" except Exception as e: output = f"错误：{{e}}" results.append({{"type": "tool_result", "tool_use_id": b.id, "content": output[:50000]}}) history.append({{"role": "user", "content": results}}) if __name__ == "__main__": h = [] print("{name} - Level 0 Agent\\n输入 q 退出。\\n") while (q := input(">> ").strip()) not in ("q", "quit", ""): print(run(q, h), "\\n") ''', 1: '''#!/usr/bin/env python3 """ Level 1 Agent - 模型即 Agent（约 200 行） 核心洞察：4 个工具覆盖 90% 的编码任务。 模型本身就是 Agent，代码只负责运行循环。 """ from anthropic import Anthropic from dotenv import load_dotenv from pathlib import Path import subprocess import os load_dotenv() client = Anthropic( api_key=os.getenv("ANTHROPIC_API_KEY"), base_url=os.getenv("ANTHROPIC_BASE_URL") ) MODEL = os.getenv("MODEL_NAME", "claude-sonnet-4-20250514") WORKDIR = Path.cwd() SYSTEM = f"""你是位于 {{WORKDIR}} 的编码 Agent。 规则： - 优先使用工具，而不是只写解释；直接行动。 - 不要编造文件路径。不确定时先使用 ls/find。 - 保持最小改动，不要过度设计。 - 完成后总结修改内容。""" TOOLS = [ {{"name": "bash", "description": "运行 shell 命令。", "input_schema": {{"type": "object", "properties": {{"command": {{"type": "string"}}}}, "required": ["command"]}}}}, {{"name": "read_file", "description": "读取文件内容。", "input_schema": {{"type": "object", "properties": {{"path": {{"type": "string"}}}}, "required": ["path"]}}}}, {{"name": "write_file", "description": "向文件写入内容。", "input_schema": {{"type": "object", "properties": {{"path": {{"type": "string"}}, "content": {{"type": "string"}}}}, "required": ["path", "content"]}}}}, {{"name": "edit_file", "description": "在文件中替换一次完全匹配的文本。", "input_schema": {{"type": "object", "properties": {{"path": {{"type": "string"}}, "old_text": {{"type": "string"}}, "new_text": {{"type": "string"}}}}, "required": ["path", "old_text", "new_text"]}}}}, ] def safe_path(p: str) -> Path: """防止路径逃逸出工作区。""" path = (WORKDIR / p).resolve() if not path.is_relative_to(WORKDIR): raise ValueError(f"路径逃逸出工作区：{{p}}") return path def execute(name: str, args: dict) -> str: """执行工具并返回结果。""" if name == "bash": dangerous = ["rm -rf /", "sudo", "shutdown", "> /dev/"] if any(d in args["command"] for d in dangerous): return "错误：已阻止危险命令" try: r = subprocess.run(args["command"], shell=True, cwd=WORKDIR, capture_output=True, text=True, timeout=60) return (r.stdout + r.stderr).strip()[:50000] or "（无输出）" except subprocess.TimeoutExpired: return "错误：执行超时（60 秒）" except Exception as e: return f"错误：{{e}}" if name == "read_file": try: return safe_path(args["path"]).read_text()[:50000] except Exception as e: return f"错误：{{e}}" if name == "write_file": try: p = safe_path(args["path"]) p.parent.mkdir(parents=True, exist_ok=True) p.write_text(args["content"]) return f"已写入 {{len(args['content'])}} 字节到 {{args['path']}}" except Exception as e: return f"错误：{{e}}" if name == "edit_file": try: p = safe_path(args["path"]) content = p.read_text() if args["old_text"] not in content: return f"错误：在 {{args['path']}} 中未找到目标文本" p.write_text(content.replace(args["old_text"], args["new_text"], 1)) return f"已编辑 {{args['path']}}" except Exception as e: return f"错误：{{e}}" return f"未知工具：{{name}}" def agent(prompt: str, history: list = None) -> str: """运行 Agent 循环。""" if history is None: history = [] history.append({{"role": "user", "content": prompt}}) while True: response = client.messages.create( model=MODEL, system=SYSTEM, messages=history, tools=TOOLS, max_tokens=8000 ) history.append({{"role": "assistant", "content": response.content}}) if response.stop_reason != "tool_use": return "".join(b.text for b in response.content if hasattr(b, "text")) results = [] for block in response.content: if block.type == "tool_use": print(f"> {{block.name}}: {{str(block.input)[:100]}}") output = execute(block.name, block.input) print(f" {{output[:100]}}...") results.append({{"type": "tool_result", "tool_use_id": block.id, "content": output}}) history.append({{"role": "user", "content": results}}) if __name__ == "__main__": print(f"{name} - Level 1 Agent，工作目录：{{WORKDIR}}") print("输入 q 退出。\\n") h = [] while True: try: query = input(">> ").strip() except (EOFError, KeyboardInterrupt): break if query in ("q", "quit", "exit", ""): break print(agent(query, h), "\\n") ''', } ENV_TEMPLATE = '''# API 配置 ANTHROPIC_API_KEY=sk-xxx ANTHROPIC_BASE_URL=https://api.anthropic.com MODEL_NAME=claude-sonnet-4-20250514 ''' def create_agent(name: str, level: int, output_dir: Path): """创建新的 Agent 项目。""" # 校验等级 if level not in TEMPLATES and level not in (2, 3, 4): print(f"错误：脚手架暂未实现 Level {level}。") print("可用等级：0（最小版）、1（4 个工具）") print("Level 2-4 请从 mini-claude-code 仓库复制。") sys.exit(1) # 创建输出目录 agent_dir = output_dir / name agent_dir.mkdir(parents=True, exist_ok=True) # 写入 Agent 文件 agent_file = agent_dir / f"{name}.py" template = TEMPLATES.get(level, TEMPLATES[1]) agent_file.write_text(template.format(name=name)) print(f"已创建：{agent_file}") # 写入 .env.example env_file = agent_dir / ".env.example" env_file.write_text(ENV_TEMPLATE) print(f"已创建：{env_file}") # 写入 .gitignore gitignore = agent_dir / ".gitignore" gitignore.write_text(".env\n__pycache__/\n*.pyc\n") print(f"已创建：{gitignore}") print(f"\nAgent '{name}' 已创建于 {agent_dir}") print(f"\n下一步：") print(f" 1. cd {agent_dir}") print(f" 2. cp .env.example .env") print(f" 3. 编辑 .env，填入你的 API key") print(f" 4. pip install anthropic python-dotenv") print(f" 5. python {name}.py") def main(): parser = argparse.ArgumentParser( description="搭建一个新的 AI 编码 Agent 项目", formatter_class=argparse.RawDescriptionHelpFormatter, epilog=""" 等级： 0 最小版（约 50 行） - 单一 bash 工具，通过自递归实现子 Agent 1 基础版（约 200 行）- 4 个核心工具：bash、read、write、edit 2 Todo（约 300 行） - 增加 TodoWrite，用于结构化规划 3 子 Agent（约 450） - 增加 Task 工具，用于上下文隔离 4 Skills（约 550） - 增加 Skill 工具，用于领域能力 """ ) parser.add_argument("name", help="要创建的 Agent 名称") parser.add_argument("--level", type=int, default=1, choices=[0, 1, 2, 3, 4], help="复杂度等级（默认：1）") parser.add_argument("--path", type=Path, default=Path.cwd(), help="输出目录（默认：当前目录）") args = parser.parse_args() create_agent(args.name, args.level, args.path) if __name__ == "__main__": main()