Loading ToolBench
MCP server that answers questions using Tavily search API
Single tool with minimal definition quality. Tool name 'tavily-question-answer' lacks action verb prefix (should be 'search_' or 'answer_'). Description is 47 chars, acceptable length but generic ('This tool is used to answer questions using Tavily'). Input schema present with single string parameter 'question' that has a description. No output schema documented. No error handling guidance beyond generic error messages. No parameter constraints (min/max length, format). Tool does not follow verb_noun naming convention. Missing documentation of what fields are returned and how to chain with other tools.
This tool is used to answer questions using Tavily
Tool name lacks action verb. 'tavily-question-answer' should start with a verb like 'search_', 'answer_', or 'query_' to signal intent to LLMs. Current name is ambiguous about what action occurs.
No output schema documented. Tool returns text content but LLMs cannot infer the structure of the response (fields, types, pagination). This prevents downstream tool chaining and forces LLMs to parse unstructured output.
Parameter 'question' lacks constraints. No min/max length, format specification, or guidance on question complexity. LLMs may pass empty strings or extremely long queries without validation.
Error handling returns generic error messages without recovery guidance. 'Error: {message}' tells LLM nothing about whether to retry, ask user, or abandon. Should include actionable next steps.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 38 | 2026-07-28+ | v2 |
Tool description is generic and does not explain when to use this tool vs alternatives, what data it returns, or prerequisites (Tavily API key required). Lacks WHAT/WHEN/RETURN structure.