Agent Skills 开发完全指南

--- name: weather-query version: 1.0.0 description: 查询全球城市天气信息 author: miaoquai license: MIT inputs: - name: city type: string required: true description: 城市名称 outputs: - name: weather_data type: object description: 天气数据 dependencies: - name: requests version: ">=2.28.0" tags: - weather - api - utility --- # 天气查询 Skill ## 使用步骤 1. 获取 API Key（从 openweathermap.org 注册） 2. 设置环境变量 WEATHER_API_KEY 3. 运行 Skill ## Python 脚本（scripts/query_weather.py） def get_weather(city): api_key = os.getenv("WEATHER_API_KEY") params = {"q": city, "appid": api_key, "units": "metric", "lang": "zh_cn"} response = requests.get("http://api.openweathermap.org/data/2.5/weather", params=params) response.raise_for_status() data = response.json() return { "city": data["name"], "temperature": data["main"]["temp"], "humidity": data["main"]["humidity"], "description": data["weather"][0]["description"] }

--- name: file-batch-processor version: 1.0.0 description: 批量处理文件（转换格式、重命名、压缩等） author: miaoquai license: MIT inputs: - name: input_dir type: string required: true description: 输入目录路径 - name: output_dir type: string required: true description: 输出目录路径 - name: operation type: string required: true description: 操作类型 enum: ["convert", "rename", "compress"] outputs: - name: processed_count type: integer description: 成功处理的文件数量 dependencies: - name: Pillow version: ">=9.0.0" tags: - file - batch - utility --- # 文件批量处理 Skill ## 使用示例 openclaw run file-batch-processor \ --input_dir ./images \ --output_dir ./converted \ --operation convert ## Python 脚本（scripts/process_files.py） def batch_convert_images(input_dir, output_dir, target_format="PNG"): import os from PIL import Image os.makedirs(output_dir, exist_ok=True) count = 0 for filename in os.listdir(input_dir): if filename.lower().endswith(('.png', '.jpg', '.jpeg')): img = Image.open(os.path.join(input_dir, filename)) name = os.path.splitext(filename)[0] img.save(os.path.join(output_dir, f"{name}.{target_format.lower()}")) count += 1 return count

--- name: mcp-integration-demo version: 1.0.0 description: 集成 MCP 服务器，演示 MCP 协议使用 author: miaoquai license: MIT inputs: - name: server_url type: string required: true description: MCP 服务器 URL - name: command type: string required: true description: 要执行的命令 outputs: - name: result type: object description: MCP 服务器返回的结果 dependencies: - name: mcp-client version: ">=1.0.0" tags: - mcp - integration - protocol --- # MCP 集成演示 Skill ## 什么是 MCP？ MCP（Model Context Protocol）是一个开放协议，允许 AI 应用 与外部数据源和工具无缝集成。 ## 使用示例 openclaw run mcp-integration-demo \ --server_url "http://localhost:8080" \ --command "list_tools" ## Python 脚本（scripts/mcp_client.py） class MCPClient: def __init__(self, server_url): self.server_url = server_url def list_tools(self): resp = requests.get(f"{self.server_url}/tools") return resp.json() def call_tool(self, name, params): resp = requests.post(f"{self.server_url}/call", json={"name": name, "parameters": params}) return resp.json() def list_resources(self): resp = requests.get(f"{self.server_url}/resources") return resp.json()