development-workflow

Development Workflow

This document outlines the mandatory workflow for all development tasks. It ensures that every action is preceded by proper analysis, follows established patterns, and utilizes the most effective tools available.

1. Mandatory Protocol: Analyze Before Action

This is the most critical rule in the workflow. It is not optional.

NEVER read, modify, or work with any file without FIRST running analyze_file. This is a non-negotiable safety and quality protocol.

1.1. The Analysis Workflow

1. Analyze the File: Before any operation, run analyze_file to get a comprehensive overview.

   {
     "tool": "analyze_file",
     "args": { "filePath": "/path/to/your/file.ts", "includeErrors": true }
   }
   

2. Assess Health: Check the health_summary.overall_status.
   • "healthy": ✅ Safe to proceed.
   • "has_warnings": ⚠️ Proceed with caution.
   • "needs_attention": ❌ STOP. You MUST fix the TypeScript errors before making any other changes.
3. Understand Context: Review the file's structure, imports, and exports to understand its role in the system.
4. Store Insights: After analysis, store key findings in the memory system to accelerate future work.

1.2. Post-Modification Verification

After every modification, you MUST re-run analyze_file to verify that no new errors have been introduced and that the file's health has been maintained or improved.

2. Tool Prioritization: Use Specialized Tools

To ensure accuracy, context-awareness, and knowledge retention, you MUST prioritize specialized MCP Server tools over generic alternatives.

Tool Selection Hierarchy

Task	✅ ALWAYS USE (Primary MCP Tools)	❌ NEVER USE (Generic Alternatives)
File Analysis	analyze_file, analyze_feature	read_file (for initial analysis)
Code Investigation	find_implementation, explore_source, trace_dependency_chain	grep_search, codebase_search
API Development	get_openapi_spec, make_api_request	curl (unless make_api_request fails)
Knowledge Storage	store_memory, relate_memories	Storing info only in conversation history
Research	get_documentation, search_github_issues	Relying on outdated training data
Task Management	create_task, list_tasks, update_task_status	Manual, ad-hoc tracking
Agent Delegation	delegate_to_agent, monitor_agent_tasks	Performing all tasks manually