security-guidelines

Security Guidelines

Input Validation

• Always validate inputs from AI models before processing
• Sanitize file paths to prevent directory traversal attacks
• Validate command parameters before executing system operations
• Check parameter types and ranges according to tool specifications

File System Operations

• Use appropriate permissions for file operations
• Restrict file access to designated directories only
• Validate file extensions and MIME types when applicable
• Never execute user-provided file paths directly

System Operations

• Be cautious with system-level operations that could affect the host
• Use subprocess safely with proper argument validation
• Limit resource usage to prevent DoS attacks
• Implement timeouts for long-running operations

Network Security

• Validate network endpoints before making connections
• Use secure protocols when possible (HTTPS, secure WebSocket)
• Implement rate limiting for resource-intensive operations
• Log security-relevant operations for audit trails

Error Handling

• Don't expose sensitive information in error messages
• Use generic error messages for security-sensitive failures
• Log detailed errors securely without exposing them to clients
• Handle exceptions gracefully to prevent information leakage

Example Secure Tool Implementation

@mcp.tool()
async def secure_file_operation(filepath: str) -> str:
    """Securely operate on files with validation."""
    # Validate input
    if not filepath or not isinstance(filepath, str):
        raise ValueError("Invalid filepath provided")
    
    # Sanitize path
    safe_path = Path(filepath).resolve()
    allowed_dir = Path("/allowed/directory").resolve()
    
    # Check if path is within allowed directory
    if not str(safe_path).startswith(str(allowed_dir)):
        raise ValueError("Access denied: path outside allowed directory")
    
    # Proceed with secure operation
    return "Operation completed safely"