Loading ToolBench
A demonstration MCP server with basic mathematical tools
Two tools with minimal definitions. Both have basic input schemas with typed parameters and descriptions, but tool descriptions are extremely short (under 50 chars) and lack context for LLM selection. No output schemas documented. No error handling guidance. No parameter constraints (enums, ranges). Tool names lack action verbs (add is acceptable, but celcius_to_fahrenheit is a noun phrase). No indication of idempotency, side effects, or when to use each tool. Descriptions are in Spanish, creating a language mismatch with English parameter names.
Suma dos números enteros y devuelve el resultado
Convierte grados Celcius a Fahrenheit
Tool descriptions are critically short (under 50 chars) and lack LLM-selection context. 'Suma dos números enteros y devuelve el resultado' (Spanish) does not explain WHEN to use this tool vs alternatives, what it returns, or prerequisites.
No output schemas documented. LLMs cannot plan downstream calls or extract return values without knowing the response structure. Both tools return simple scalars but this is not formally declared.
Tool name 'celcius_to_fahrenheit' is a noun phrase, not a verb_noun pattern. Should be 'convert_celsius_to_fahrenheit' to signal the action clearly to LLMs.
Descriptions are in Spanish while parameter names and code are in English. This language mismatch creates confusion for LLMs and violates consistency expectations.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 38 | 2026-07-28+ | v2 |
No error handling guidance. If add() receives non-integer inputs or celcius_to_fahrenheit() receives invalid temperature values, there is no recovery path documented for the LLM.