Structured Workflow MCP

# MCP Server Testing Notes ## Test Evaluation Criteria This document outlines what to observe when testing the Structured Workflow MCP Server with the refactor-test prompt. ### 1. Tool Discovery and Usage **Expected Behavior:** - AI should discover and use MCP workflow tools starting with `mcp__structured-workflow__refactor_workflow` - AI should follow the phase progression naturally - AI should request guidance for each phase **What to Watch For:** - Does the AI find and use the MCP tools? - Does it follow the suggested workflow? - Does it skip phases or try to jump ahead? ### 2. Phase Execution Quality #### AUDIT Phase - Should read files without modifying them - Should identify all major code smells and SOLID violations - Should document findings comprehensively #### INVENTORY Phase - Should create a detailed list of all violations - Should categorize issues by SOLID principle - Should prioritize refactoring tasks #### COMPARE/ANALYZE Phase - Should consider multiple refactoring approaches - Should evaluate trade-offs - Should justify chosen approach #### WRITE/REFACTOR Phase - Should implement clean, separated classes - Should use proper abstractions - Should maintain all functionality ### 3. SOLID Violations to Identify The test UserService contains these violations: **Single Responsibility Violations:** - User management - Authentication - Logging - Email service - Notification system - Data export/import - Report generation - Database operations - Utility functions **Open/Closed Violations:** - Hardcoded email configuration - Hardcoded export formats - Hardcoded report templates **Liskov Substitution Violations:** - `getUserById` method with inconsistent behavior **Interface Segregation Violations:** - Admin functions mixed with user functions - Unrelated utility methods **Dependency Inversion Violations:** - Direct dependency on fs module - Direct dependency on nodemailer - Hardcoded database operations ### 4. Expected Refactoring Outcomes A successful refactoring should produce: 1. **Separate Services/Classes:** - UserRepository (data access) - AuthenticationService - Logger/ILogger interface - EmailService/IEmailService interface - NotificationService - DataExportService - ReportGenerator - PasswordHasher 2. **Proper Abstractions:** - Interfaces for all services - Dependency injection in constructors - Configuration objects instead of hardcoding 3. **Clean Architecture:** - Clear separation of concerns - Testable components - Extensible design ### 5. Red Flags Watch for these anti-patterns: - Skipping the AUDIT phase and jumping to refactoring - Not using the MCP workflow tools - Modifying the original files instead of creating new ones - Incomplete refactoring (fixing some but not all violations) - Breaking existing functionality ### 6. Success Metrics Rate the MCP server's performance on: 1. **Guidance Quality** (1-5): How helpful were the phase instructions? 2. **Workflow Flow** (1-5): How smooth was the phase progression? 3. **Violation Detection** (1-5): How well did it help identify issues? 4. **Refactoring Quality** (1-5): How good was the final refactored code? 5. **User Experience** (1-5): How easy was it to follow the guidance? ### 7. Testing Variations Consider testing with: - Different AI models (Claude, Cursor, Windsurf) - Skipping phases intentionally - Trying to modify files before reading - Requesting help mid-workflow - Starting over after partial completion ## Reporting Issues If the MCP server fails to guide properly, note: - Which phase failed - What guidance was missing or unclear - What the AI did instead - How it could be improved This test is designed to validate that the MCP server successfully guides AI assistants through professional refactoring workflows while maintaining AI autonomy and tool flexibility.