Loading ToolBench
Let the scripts your coding agent writes call your MCP servers — no extra auth. Code mode for the agent you already run.
Single tool 'echo' with minimal definition. Tool has a basic description (18 chars, below 20-char threshold) and a properly typed input schema. However, the description is too terse to guide LLM selection ('Echo the message back.' lacks context on when/why to use it). No output schema documented. No error handling guidance. No parameter constraints or validation hints. This is a test/demo server, not production-grade.
Echo the message back.
Tool description too short (18 chars). Below 20-char minimum; cannot guide LLM selection. Expand to 50-200 chars explaining what the tool does, when to use it, and what it returns.
No output schema documented. LLMs cannot plan downstream calls or extract structured data without knowing the response format. Document the return type (e.g., {type: 'object', properties: {text: {type: 'string'}}}).
Tool name 'echo' is generic and does not follow verb_noun convention. Rename to 'echo_message' or 'repeat_text' to clarify intent and match production naming baselines (90% of A+ tools start with action verb).
No error handling guidance. If the message parameter is empty or exceeds a size limit, what should the LLM do? Provide recovery hints in error responses.
Inferred effective spec: 2025-06-18+.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 38 | 2025-06-18+ | v2 |
Parameter 'message' description is minimal ('text to echo'). Expand to clarify format constraints, max length, and any special handling (e.g., 'Plain text string, max 10000 chars. Newlines preserved.').