SlopWatch MCP Server

Instructions

Implementation Reference

• src/mcp-server.js:319-437 (handler)

  The handleSetupRules function implements the core logic of the 'slopwatch_setup_rules' tool. It extracts the project_path and overwrite parameters, generates a hardcoded multi-line string containing the .cursorrules content for SlopWatch AI accountability rules (version 2.6.0), and returns it as a formatted text response block.

    async handleSetupRules(args) {
      const { project_path, overwrite = false } = args;
      
      // Return updated rules content for v2.6.0 with combined tool and minimal responses
      const rulesContent = `# SlopWatch AI Accountability Rules v2.6.0
  
  ## MANDATORY PROTOCOL
  
  ### OPTION 1: COMBINED TOOL (RECOMMENDED ⭐)
  For most implementations, use the single-call approach:
  \`\`\`
  mcp_SlopWatch_slopwatch_claim_and_verify({
    claim: "What you implemented",
    originalFileContents: { "filename": "before_content" },
    updatedFileContents: { "filename": "after_content" }
  });
  // Response: "✅ PASSED (87%)" or "❌ FAILED (23%)"
  \`\`\`
  
  ### OPTION 2: TRADITIONAL 2-STEP (WHEN NEEDED)
  For pre-planning implementations:
  \`\`\`
  // Step 1: Before implementing
  mcp_SlopWatch_slopwatch_claim({
    claim: "What you're about to implement",
    fileContents: { "filename": "current_content" }
  });
  // Response: "Claim ID: abc123"
  
  // Step 2: After implementing  
  mcp_SlopWatch_slopwatch_verify({
    claimId: "abc123",
    updatedFileContents: { "filename": "updated_content" }
  });
  // Response: "✅ PASSED (92%)" or "❌ FAILED (15%)"
  \`\`\`
  
  ## TRIGGER PHRASES
  When you mention ANY of these, you MUST use SlopWatch:
  - "I will implement/add/create/modify/fix/update/build/write"
  - "Let me implement/add/create/modify/fix/update"
  - "I'll add/create/implement/modify/fix/update"
  
  ## ENFORCEMENT RULES
  
  ### NEVER SKIP VERIFICATION
  - Every implementation claim requires verification
  - No exceptions for "simple" changes
  - Failed verification = STOP and fix immediately
  
  ### ULTRA-MINIMAL RESPONSES (v2.6.0)
  SlopWatch now returns minimal responses:
  - Claim: "Claim ID: abc123"
  - Verify: "✅ PASSED (87%)" or "❌ FAILED (23%)"
  - Combined: "✅ PASSED (87%)" or "❌ FAILED (23%)"
  - Status: "Accuracy: 95% (19/20)"
  
  ### WORKFLOW PATTERNS
  
  #### RECOMMENDED: Combined Tool Workflow
  \`\`\`
  I'll implement [feature]. Let me verify it with SlopWatch.
  
  [Implement the code changes]
  
  [Call mcp_SlopWatch_slopwatch_claim_and_verify with before/after content]
  
  ✅ PASSED (87%) - Implementation verified!
  \`\`\`
  
  #### ALTERNATIVE: Traditional Workflow  
  \`\`\`
  I'll implement [feature]. Let me register this with SlopWatch first.
  
  [Call mcp_SlopWatch_slopwatch_claim]
  Claim ID: abc123
  
  [Implement the code changes]
  
  [Call mcp_SlopWatch_slopwatch_verify]
  ✅ PASSED (92%) - Implementation verified!
  \`\`\`
  
  #### FAILURE HANDLING
  \`\`\`
  ❌ FAILED (23%) - SlopWatch verification failed.
  Let me analyze and fix the implementation.
  [Fix the code and verify again]
  \`\`\`
  
  ## SPECIAL CASES
  
  ### NO CLAIM NEEDED:
  - Reading/analyzing code
  - Explaining existing code  
  - Answering questions
  - Code reviews
  
  ### REQUIRES CLAIMS:
  - Creating/modifying files
  - Adding functions/classes
  - Configuration changes
  - Package installations
  
  ## EMERGENCY BYPASS
  Only if SlopWatch is unavailable:
  "⚠️ SlopWatch unavailable - proceeding without verification"
  
  Remember: SlopWatch v2.6.0 features ultra-minimal responses and combined tools for seamless AI accountability.`;
  
      return {
        content: [
          {
            type: 'text',
            text: rulesContent
          }
        ]
      };
    }

• src/mcp-server.js:80-98 (schema)

  The input schema definition for the 'slopwatch_setup_rules' tool, specifying required 'project_path' (string) and optional 'overwrite' (boolean, default false). This is returned in the ListTools response.

  {
    name: 'slopwatch_setup_rules',
    description: 'Generate .cursorrules file with AI accountability enforcement',
    inputSchema: {
      type: 'object',
      properties: {
        project_path: {
          type: 'string',
          description: 'Path to project directory where .cursorrules should be created'
        },
        overwrite: {
          type: 'boolean',
          description: 'Whether to overwrite existing .cursorrules file',
          default: false
        }
      },
      required: ['project_path']
    }
  }

• src/mcp-server.js:104-117 (registration)

  The tool call handler registration via setRequestHandler(CallToolRequestSchema). The switch statement routes 'slopwatch_setup_rules' calls to the handleSetupRules method.

  this.server.setRequestHandler(CallToolRequestSchema, async (request) => {
    const { name, arguments: args } = request.params;
  
    switch (name) {
      case 'slopwatch_claim_and_verify':
        return await this.handleClaimAndVerify(args);
      case 'slopwatch_status':
        return await this.handleStatus(args);
      case 'slopwatch_setup_rules':
        return await this.handleSetupRules(args);
      default:
        throw new Error(`Unknown tool: ${name}`);
    }
  });