AgentSkillsCN

Add MCP Tool

添加MCP工具

SKILL.md

Add MCP Tool

Create a new tool for the MCP (Model Context Protocol) server.

Description

Adds a new tool to the MCP server that AI agents can discover and invoke. Tools provide structured access to data and functionality.

Instructions

  1. •Define the tool in lib/mcp-tools.ts or create a new module in lib/mcp/
  2. •Create the tool object with name, description, inputSchema, and handler
  3. •Add to the mcpTools export object
  4. •Tool will auto-register when the server restarts

Parameters

  • •tool_name - Snake_case identifier for the tool
  • •description - Clear description of what the tool does
  • •input_params - Parameters the tool accepts
  • •return_type - What the tool returns

Template

typescript
// In lib/mcp-tools.ts or lib/mcp/[module]/index.ts

export const my_new_tool = {
  name: 'my_new_tool',
  description: 'Clear description of what this tool does and when to use it',
  inputSchema: {
    type: 'object',
    properties: {
      required_param: {
        type: 'string',
        description: 'Description of this parameter'
      },
      optional_param: {
        type: 'number',
        default: 10,
        description: 'Optional parameter with default'
      },
      enum_param: {
        type: 'string',
        enum: ['option1', 'option2', 'option3'],
        description: 'Parameter with fixed options'
      }
    },
    required: ['required_param']
  },
  handler: async ({ required_param, optional_param = 10, enum_param }: {
    required_param: string;
    optional_param?: number;
    enum_param?: string;
  }) => {
    try {
      // Implementation logic
      const result = await someOperation(required_param);
      
      return {
        content: [
          {
            type: 'text',
            text: `Successfully processed: ${result}`
          }
        ]
      };
    } catch (error) {
      console.error('Tool error:', error);
      return {
        content: [
          {
            type: 'text',
            text: `Error: ${error instanceof Error ? error.message : 'Unknown error'}`
          }
        ]
      };
    }
  }
};

Adding to mcpTools

typescript
// In lib/mcp-tools.ts
export const mcpTools = {
  // ... existing tools
  my_new_tool,
};

Best Practices

  1. •Use snake_case for tool names (MCP convention)
  2. •Write clear descriptions - AI agents use these to decide when to use tools
  3. •Validate input - Even though schema provides validation, check edge cases
  4. •Return structured content - Use the content array format
  5. •Handle errors gracefully - Return error messages, don't throw
  6. •Log errors - Use console.error for debugging