mcp-prompt-implementation

MCP Prompt Implementation Guidelines

When implementing new Model Context Protocol (MCP) prompts for the Unleash MCP Server, follow these patterns:

Structure

Each MCP prompt file should contain:

1. A parameter schema using Zod for validation
2. A handler function that processes inputs and returns formatted messages
3. An exported prompt definition object

Naming Convention

• File name: feature-name.ts (kebab-case)
• Schema: FeatureNameParamsSchema (PascalCase + ParamsSchema suffix)
• Handler: handleFeatureNamePrompt (handle + PascalCase + Prompt suffix)
• Export: featureNamePrompt (camelCase + Prompt suffix)

Implementation Template

/**
 * Brief description of the prompt purpose
 */

import { z } from 'zod';

/**
 * Define any needed schemas for validation
 */
const RequiredDataSchema = z.object({
  // Add properties with appropriate Zod validators
  property: z.string(),
  optionalProperty: z.number().optional()
}).passthrough();

/**
 * Parameters schema for this prompt
 */
export const FeatureNameParamsSchema = {
  requiredParam: z.string(),
  optionalData: RequiredDataSchema.optional()
};

/**
 * Handler for this prompt
 */
export function handleFeatureNamePrompt({ 
  requiredParam, 
  optionalData = {} 
}: { 
  requiredParam: string; 
  optionalData?: any 
}) {
  return {
    messages: [{
      role: "user",
      content: {
        type: "text",
        text: `Prompt text that uses ${requiredParam} and:
${JSON.stringify(optionalData, null, 2)}

Instructions for the AI on how to respond, including which tools to use.`
      }
    }]
  };
}

/**
 * Prompt definition export
 */
export const featureNamePrompt = {
  name: "descriptiveName",
  paramsSchema: FeatureNameParamsSchema,
  handler: handleFeatureNamePrompt
};

Best Practices

1. Use Zod for all parameter validation
2. Include clear JSDoc comments
3. Format multi-line strings with template literals
4. Provide default values for optional parameters
5. Use proper TypeScript typing for handler parameters
6. Format JSON with indentation for readability
7. Include instructions for tool usage in the prompt text
8. Export only the final prompt definition object
9. Follow the project structure for importing dependencies