PromptTuner MCP

// Quick Workflow Prompts for PromptTuner MCP import type { McpServer } from '@modelcontextprotocol/sdk/server/mcp.js'; import { z } from 'zod'; export function registerQuickWorkflowPrompts(server: McpServer): void { // Quick Optimize - single technique, fast server.registerPrompt( 'quick-optimize', { title: 'Quick Optimize', description: 'Fast prompt improvement with grammar and clarity fixes.', argsSchema: { prompt: z.string().min(1).describe('The prompt to optimize'), }, }, ({ prompt }) => ({ messages: [ { role: 'user', content: { type: 'text', text: `Use refine_prompt with technique "basic" on this prompt. Show the improved version and briefly note key changes. Prompt: ${prompt}`, }, }, ], }) ); // Deep Optimize - comprehensive, thorough server.registerPrompt( 'deep-optimize', { title: 'Deep Optimize', description: 'Comprehensive optimization with all techniques applied.', argsSchema: { prompt: z.string().min(1).describe('The prompt to optimize'), }, }, ({ prompt }) => ({ messages: [ { role: 'user', content: { type: 'text', text: `Use optimize_prompt with techniques ["comprehensive"]. Show: before/after scores, all improvements made, final optimized prompt. Prompt: ${prompt}`, }, }, ], }) ); // Full Analysis - scoring and recommendations server.registerPrompt( 'analyze', { title: 'Analyze Prompt', description: 'Score prompt quality and get improvement suggestions.', argsSchema: { prompt: z.string().min(1).describe('The prompt to analyze'), }, }, ({ prompt }) => ({ messages: [ { role: 'user', content: { type: 'text', text: `Analyze this prompt: 1. Use analyze_prompt for scores (clarity, specificity, completeness, structure, effectiveness) 2. Use detect_format to identify target format 3. Summarize: overall score, strengths, top 3 recommendations Prompt: ${prompt}`, }, }, ], }) ); // Best Practices Review server.registerPrompt( 'review', { title: 'Best Practices Review', description: 'Check prompt against prompting best practices.', argsSchema: { prompt: z.string().min(1).describe('The prompt to review'), }, }, ({ prompt }) => ({ messages: [ { role: 'user', content: { type: 'text', text: `Review this prompt against best practices: 1. Use analyze_prompt for current state 2. Check: clarity, context/role, structure, output format, constraints 3. For each gap: what's missing, why it matters, how to fix Prompt: ${prompt}`, }, }, ], }) ); // Iterative Refinement server.registerPrompt( 'iterative-refine', { title: 'Iterative Refinement', description: 'Identify top 3 weaknesses, explain each, and apply fixes iteratively.', argsSchema: { prompt: z.string().min(1).describe('The prompt to refine'), }, }, ({ prompt }) => ({ messages: [ { role: 'user', content: { type: 'text', text: `Perform iterative refinement on this prompt: 1. Use analyze_prompt to identify issues 2. Rank the top 3 weaknesses by severity 3. For each weakness: - What's wrong (specific issue) - Why it matters (impact on AI understanding) - Specific fix (concrete improvement) 4. Use optimize_prompt with appropriate techniques to apply all fixes 5. Show the final improved prompt with a summary of changes Prompt: ${prompt}`, }, }, ], }) ); // Technique Recommendation server.registerPrompt( 'recommend-techniques', { title: 'Recommend Techniques', description: 'Recommend best optimization techniques based on prompt and task type.', argsSchema: { prompt: z.string().min(1).describe('The prompt to analyze'), taskType: z .enum([ 'classification', 'analysis', 'generation', 'extraction', 'debugging', 'translation', 'summarization', 'other', ]) .optional() .default('other') .describe('Type of task the prompt is for'), }, }, ({ prompt, taskType }) => ({ messages: [ { role: 'user', content: { type: 'text', text: `Recommend optimization techniques for this prompt: Task Type: ${taskType} Process: 1. Use analyze_prompt to understand current state and weaknesses 2. Use detect_format to understand target format preference 3. Based on task type and analysis, recommend techniques in priority order: - basic: Always beneficial for grammar/clarity - chainOfThought: For reasoning, math, debugging, analysis tasks - fewShot: For classification, translation, pattern-based tasks - roleBased: When domain expertise would improve output - structured: For complex multi-part instructions - comprehensive: When prompt needs significant improvement 4. For each recommended technique: - Why it's beneficial for this prompt - Expected improvement area (clarity, structure, effectiveness) - Priority level (high/medium/low) 5. Suggest optimal technique combination for optimize_prompt Prompt: ${prompt}`, }, }, ], }) ); // Anti-Pattern Scanner server.registerPrompt( 'scan-antipatterns', { title: 'Scan Anti-Patterns', description: 'Detect common prompt anti-patterns and provide corrections.', argsSchema: { prompt: z.string().min(1).describe('The prompt to scan'), }, }, ({ prompt }) => ({ messages: [ { role: 'user', content: { type: 'text', text: `Scan this prompt for common anti-patterns: Use analyze_prompt to check for these issues: **Anti-Patterns to Detect:** 1. Vague language (something, stuff, things, etc.) 2. Missing role/persona context 3. Unclear or missing output format specification 4. No constraints or boundaries (ALWAYS/NEVER rules) 5. Lack of examples for complex/ambiguous tasks 6. Overly long run-on sentences (>30 words) 7. Ambiguous pronouns (it, this, that without clear referent) 8. Missing context for technical or domain-specific terms 9. No success criteria or quality indicators 10. Conflicting or contradictory instructions For each anti-pattern found: - Quote the problematic text - Explain why it's problematic - Provide a corrected version - Note the severity (high/medium/low impact) Then use refine_prompt with technique "comprehensive" to show a fully corrected version. Summary format: ## Anti-Patterns Detected: X [List each with severity] ## Corrected Prompt [Show improved version] ## Impact Expected improvement in clarity, specificity, and effectiveness. Prompt: ${prompt}`, }, }, ], }) ); }