Loading ToolBench
Example MCP Server that doubles numbers
Single tool with minimal documentation. Tool has explicit registration with name and Zod schema, but description is in Japanese, lacks English context for LLM selection, and parameter description is minimal (4 characters). No error handling, no output schema documentation, no composition guidance. Parameter count is 1, schema is present but description quality is extremely low by production standards.
与えられた数値を2倍にする
Description is in Japanese ('与えられた数値を2倍にする'), LLM selection relies on English descriptions. Japanese descriptions severely limit discoverability in English-language LLM agents.
Parameter 'num' has only 2-character description ('数値' = 'number' in Japanese). Parameter lacks English description, format range specification, and context for LLM selection.
No output schema documented. LLM cannot predict return structure or plan downstream operations. Tool returns text, but format is undocumented.
No error handling. Tool may fail with invalid input (null, Infinity, NaN, very large numbers) with no guidance for LLM recovery.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | F | 37 | 2026-07-28+ | v2 |
| 2026-03-09 | F | 31 | - | v1 |
Tool name 'double_number' does not follow verb_noun pattern. 'double' is imperative, not action-descriptive. Clearer names: 'multiply_number' or 'scale_number'.