Loading ToolBench
Three tools with basic schemas and descriptions, but significant gaps in parameter documentation and output schema clarity. All tools have descriptions (10-50 chars), which is below the 50-200 char baseline for LLM-optimized descriptions. Input schemas use Zod but lack detailed parameter descriptions beyond the symbol/amount fields. No output schemas documented. Error handling returns generic text responses without recovery guidance. Tool names follow verb_noun pattern (get-*, calculate-*) which is good, but descriptions are minimal and don't explain WHEN to use each tool or what the LLM should expect in responses.
Calculate the value of a given amount of a base token in the quote token
Get all trading pairs
Get price for a trading pair
Output schemas not documented. Tools return JSON text blobs without declaring field structure, forcing LLMs to parse unstructured responses.
Tool descriptions are too brief (10-50 chars vs 50-200 char baseline). 'Get price for a trading pair' and 'Get all trading pairs' lack context on WHEN to use each tool and what data structure is returned.
Error responses are generic text ('Error fetching price: [error]') without recovery guidance. LLMs cannot determine if the error is retryable, user-fixable, or fatal.
get-all-trading-pairs has no input parameters documented in schema (empty {} object). Parameter descriptions are required even for zero-parameter tools to explain what they return.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 50 | 2026-07-28+ | v2 |
No pagination support. get-all-trading-pairs returns all markets without limit or offset parameters, risking context window exhaustion if the API returns hundreds of items.