---

name: fastmcp-quality-review description: Review and score FastMCP MCP server quality. Use when creating/auditing MCP servers, before publishing to MCP directory, or to improve existing servers. triggers:

English

• is my MCP server good?
• check my server quality
• review my MCP
• how can I improve my MCP server?
• analyze the quality of this server
• give me feedback on my MCP
• how well-written is this server?

French

• mon serveur MCP est-il bon ?
• vérifie la qualité de mon serveur
• analyse mon MCP
• comment puis-je améliorer mon serveur MCP ?
• donne-moi ton avis sur mon MCP
• ce serveur est-il bien écrit ?

---

FastMCP Quality Review

Automated quality review for FastMCP MCP servers based on Anthropic MCP Directory Policy.

Quick Start

Observability

First: At the start of execution, display:

Copy🔧 Skill "fastmcp-quality-review" activated

1. Locate the MCP server file (usually mcp_server.py or similar)
2. Run the review process below
3. Get a score out of 100 with actionable recommendations

Review Process

Step 1: Load Server Code

Copy# Read the MCP server file
from fastmcp import Client
# Import the server instance

Step 2: Run Quality Checks

Execute each check from the Quality Checklist:

Step 3: Generate Report

Output format:

Copy## MCP Server Quality Report

**Server**: {server_name}
**Score**: {total}/100 ({grade})

### Results by Category

| Category | Score | Status | Issues |
|----------|-------|--------|--------|
| Tool Descriptions | X/20 | ✅/⚠️/❌ | ... |
| Error Handling | X/15 | ... | ... |
| ... | ... | ... | ... |

### Critical Issues (Must Fix)
1. ...

### Recommendations (Should Fix)
1. ...

### Nice to Have
1. ...

Scoring Guide

Review Script

Copy# In-memory review using FastMCP Client
import asyncio
from fastmcp import Client

async def review_mcp_server(mcp_server):
    """Review an MCP server and return quality score."""
    results = {
        "tool_descriptions": {"score": 0, "max": 20, "issues": []},
        "error_handling": {"score": 0, "max": 15, "issues": []},
        "annotations": {"score": 0, "max": 15, "issues": []},
        "token_efficiency": {"score": 0, "max": 15, "issues": []},
        "test_coverage": {"score": 0, "max": 20, "issues": []},
        "documentation": {"score": 0, "max": 15, "issues": []},
    }

    async with Client(transport=mcp_server) as client:
        tools = await client.list_tools()

        for tool in tools:
            # Check description length
            if len(tool.name) > 64:
                results["tool_descriptions"]["issues"].append(
                    f"Tool '{tool.name}' name exceeds 64 chars"
                )

            # Check description exists and is clear
            if not tool.description or len(tool.description) < 10:
                results["tool_descriptions"]["issues"].append(
                    f"Tool '{tool.name}' has missing/short description"
                )

            # Check for annotations (if available in schema)
            # ...

    return results

What to Check Manually

Some checks require manual review:

1. Functionality Match: Do tools actually do what descriptions say?
2. Security: OAuth 2.0 for remote servers? No hardcoded secrets?
3. Edge Cases: Test with invalid inputs
4. Token Bloat: Are responses unnecessarily verbose?

See references/manual-review-guide.md for detailed manual checks.

Skill Chaining

Skills Required Before

• None (entry point for MCP review)

Input Expected

• Path to MCP server file or module
• Optional: path to test directory

Output Produced

• Format: Markdown quality report with score
• Side effects: None (read-only analysis)

Compatible Skills After

• skill-creator: If major rewrite needed
• commit: After fixing issues

Tools Used

• Read (read server source code)
• Bash (run pytest, import checks)
• Grep (find patterns in code)