Loading ToolBench
Severe quality gaps across all dimensions. Tool descriptions are minimal (7-40 chars, well below 50-200 char baseline for A+ tools). Parameter descriptions are entirely absent (null values in schema). No output schemas documented. The 'add' tool is a toy example that dilutes focus. 'place-order' is IRREVERSIBLE but lacks confirmation/dry-run pattern. Error handling is absent, no recovery guidance, no categorization. Naming is acceptable (verb-noun pattern) but insufficient to overcome missing documentation.
Addition tool
Retrieves order history from Zerodha
Places a trading order on Zerodha
All parameter descriptions are null. LLMs cannot infer what 'a', 'b', 'tradingSymbol', 'transactionType', 'quantity' control without explicit descriptions. Violates pattern:tool-description baseline (100% of A+ tools have param descriptions).
No output schemas documented. LLMs cannot plan downstream calls or extract structured data. All tools return generic text content without field definitions. Violates pattern:tool baseline (100% of A+ tools have documented return types).
place-order is IRREVERSIBLE but lacks confirmation/dry-run pattern. Agents can execute trades without verification, risking financial loss. No error handling to guide recovery from failed orders.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 36 | 2026-07-28+ | v2 |
Tool descriptions are too short (7-40 chars vs 50-200 char baseline). 'Retrieves order history from Zerodha' and 'Places a trading order on Zerodha' lack context for WHEN to use, WHAT it returns, and prerequisites. Insufficient for LLM selection.
No error handling or recovery guidance. trade.controllers.ts catches errors but returns raw error messages or undefined. LLMs receive no actionable next steps (retry? ask user? fatal?). Violates pattern:recovery-guide.