Bridge any MCP server (stdio / SSE / StreamableHTTP) to a Xiaozhi server over WebSocket.
Single tool (calculator) with minimal definition quality. Tool has a basic description (67 chars) and one parameter with type but lacks parameter description. No output schema documented. No error handling guidance. No tool annotations. The description is functional but below the 194-char baseline for production tools. Parameter 'python_expression' has type 'string' but no description explaining format, constraints, or examples of valid expressions.
Evaluate a Python math expression. You may use `math` and `random` directly without importing them.
Parameter lacks description. 'python_expression' has type but no description explaining what expressions are valid, constraints, or format.
No output schema documented. Tool returns a dict but LLM cannot see the structure (success, result fields) without reading source code.
No error handling guidance. Tool uses eval() which can raise exceptions (SyntaxError, ZeroDivisionError, etc.) but no error recovery hints provided to LLM.
Tool name 'calculator' is generic and does not start with action verb. 'evaluate_expression' or 'compute_math' would be clearer.
No security guidance. Tool uses eval() which is a code injection risk. No mention of sandboxing, allowed modules, or input validation.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 46 | <=2025-11-25 | v2 |