AI Agent Workflow Starter Kit

You are an expert AI Detective & Code Auditor.
Analyze the provided code/error trace with a rigorous 4-step investigation:
1. Root Cause Analysis: Pinpoint the exact line and mechanism causing the failure.
2. Missing Constraints: Identify unhandled edge cases, boundary conditions, or race conditions.
3. Minimal Robust Fix: Provide the exact code replacement with minimal side effects.
4. Verification Checklist: Provide 2-3 specific test cases to prove the bug is eliminated.

Do not guess without evidence. Explain the logic clearly before presenting the code.

[User Request / User Feedback]
       │
       ▼
┌──────────────────┐
│   Orchestrator   │  (วางแผน & มอบหมายงาน)
└────────┬─────────┘
         │
         ├───► [1. Researcher Agent] : ค้นหา Codebase / Doc / Web Reference
         │
         ├───► [2. Coder Agent]      : เขียนฟังก์ชันและ Unit Tests
         │
         └───► [3. Evaluator Agent]  : รัน Test / ตรวจ Type & Linter
                    │
                    ├── (พบข้อผิดพลาด) ──► ส่ง Feedback กลับไป Coder แก้ไขอัตโนมัติ
                    └── (ผ่านเกณฑ์ 100%) ─► ส่งมอบโค้ด / สร้าง Pull Request

{
  "mcpServers": {
    "filesystem": {
      "command": "npx",
      "args": [
        "-y",
        "@modelcontextprotocol/server-filesystem",
        "/Users/username/projects/my-app"
      ]
    }
  }
}

# Senior Dev Anti-Bloat & Token Economy Ruleset
## 1. Core Engineering Principles
- YAGNI First: Build only what is asked. Do not guess future abstractions.
- Native Platform First: Prefer built-in language/runtime APIs before adding packages.
- Anti-Bloat: Do not create wrappers or utility layers unless reused 3+ times.
- Minimal Invasive Surgery: Make targeted line-range edits. Never re-write entire files.

## 2. Token Economy & Context Management
- Grep & AST First: Use ripgrep or symbol definitions before ingesting entire files.
- Never Cat Entire Repositories: Read only relevant function signatures and line slices.

## 3. Verification Loop
- Diagnosis Before Code: Prove the root cause before editing code.
- Verification Requirement: Every edit must pass syntax checks, linter, and tests.

#!/usr/bin/env python3
import os
from mcp.server.fastmcp import FastMCP

mcp = FastMCP("code-detective-agent-tools")

@mcp.tool()
def inspect_code_health(filepath: str) -> dict:
    """Analyze file size, line count, and estimate token cost for LLM."""
    if not os.path.exists(filepath):
        return {"error": f"File not found: {filepath}"}
    
    with open(filepath, "r", encoding="utf-8", errors="ignore") as f:
        lines = f.readlines()
    
    char_count = sum(len(line) for line in lines)
    est_tokens = char_count // 4
    
    return {
        "status": "success",
        "filepath": filepath,
        "total_lines": len(lines),
        "total_characters": char_count,
        "estimated_tokens": est_tokens,
        "recommendation": "Lean & Clean" if est_tokens < 2000 else "Needs Chunking or Slicing"
    }

if __name__ == "__main__":
    mcp.run()