AgentSkillsCN

mcp-todo-skill

【做什么】创建一个无状态的 MCP 服务器，用于待办事项管理。 【何时】当用户希望通过 AI 管理待办事项、为任务搭建 MCP 服务器，或为对话式智能体实现“待办事项”工具集时使用。

SKILL.md
--- frontmatter
name: mcp-todo-skill
description: |
  [What] Creates a stateless MCP server for todo management.
  [When] Use when users ask to manage todos via AI, set up an MCP server for tasks, or implement the 'todo' toolset for conversational agents.
allowed-tools: Write, Read

MCP Todo Skill

Creates a production-ready, stateless Model Context Protocol (MCP) server for managing todo tasks.

What This Skill Does

  • •Generates a fully functional Python MCP server using FastMCP.
  • •Implements 5 core tools: add, list, complete, delete, update.
  • •Enforces strict user isolation via user_id arguments.
  • •Persists data to PostgreSQL.
  • •Provides instructions for running and registering the server.

What This Skill Does NOT Do

  • •Handle user authentication (assumes user_id is provided by trusted caller).
  • •Create the database schema (assumes existing or external setup).
  • •Deploy the server (provides local execution instructions only).

Before Implementation

Gather context to ensure successful implementation:

SourceGather
CodebaseDB connection details (DATABASE_URL), existing SQLModel definitions.
ConversationSpecific user needs (custom fields, extra status types?).
Skill Referencesreferences/mcp-concepts.md (server setup), references/user-isolation.md (security).
User GuidelinesProject naming conventions, architectural constraints (statelessness).

Ensure all required context is gathered before implementing. Only ask user for THEIR specific requirements (domain expertise is in this skill).

Implementation Workflow

  1. •

    Verify Prerequisites:

    • •Python 3.10+
    • •mcp SDK installed
    • •sqlmodel and psycopg2-binary (or asyncpg) installed
  2. •

    Generate Server Code:

    • •Use assets/server_template.py as the base.
    • •Adjust DATABASE_URL and TodoTask model definition to match backend/models.py if available.
  3. •

    Register Extension:

    • •Create a claude_mcp_config.json entry or update .claude/mcp-config.json.
    • •Command: uvicorn mcp_server:mcp.app or python mcp_server.py.
  4. •

    Verify Tools:

    • •Ensure all 5 tools are exposed.
    • •Check error handling (try accessing non-existent task).

Output Checklist

  • • Server instantiates FastMCP("Todo Manager").
  • • 5 tools implemented: add_task, list_tasks, complete_task, delete_task, update_task.
  • • All tools accept user_id as first argument.
  • • No global state or in-memory storage used.
  • • Returns strictly typed dictionaries/JSON.

Reference Files

FileWhen to Read
references/mcp-concepts.mdFor FastMCP specifics & error handling patterns.
references/user-isolation.mdFor security & stateless design rules.
assets/server_template.pySource code for the MCP server implementation.