Single tool with minimal definition. check_connection has a basic description (31 chars) but empty input schema {}. The server dynamically generates tools from an IDA plugin file (mcp-plugin.py) via AST parsing, but the source code provided does not show the actual plugin file or generated tool definitions. Tool registration is inferred from the MCPVisitor class and code generation logic, not directly visible. No parameter descriptions, no output schema documentation, no error handling guidance. The architecture is clever but opaque to evaluation, generated tools cannot be scored without seeing mcp-plugin.py.

• Include mcp-plugin.py in the source code submission so generated tool definitions can be evaluated. Without it, schemas and descriptions are invisible.
• Add explicit JSON Schema definitions to check_connection input (currently empty {}). Even if no parameters are needed, document this explicitly: 'input: {"type": "object", "properties": {}, "required": []}'.
• Document the output schema for check_connection and all generated tools. E.g., 'Returns: {"type": "string", "description": "Connection status message"}' or structured object with typed fields.
• Add parameter descriptions to generated tools via the Annotated[type, Field(description=...)] pattern already used in MCPVisitor. Ensure every parameter has a non-empty description.
• Enhance error handling with recovery guidance. When JSON-RPC fails, return actionable messages like 'IDA plugin not running. Start it via Edit → Plugins → MCP (Ctrl+Alt+M) and retry.' for all tools, not just check_connection.

Inferred effective spec: <=2025-11-25.

• Relies on Logging (deprecated) - log to stderr or use OpenTelemetry