Loading ToolBench
Single tool with adequate schema but weak descriptions and no error guidance. Tool name 'evaluate_mathematica' is clear and verb-prefixed, but description (127 chars) lacks context on when to use it, prerequisites, or what happens on failure. Input schema properly typed but parameter description is generic. No output schema documented. Error handling returns strings instead of structured errors, forcing LLM to parse text. No pagination, no idempotency guarantees, no security considerations for arbitrary code execution.
Evaluate a Mathematica expression using wolframscript. Args: expression: The Mathematica expression string to evaluate. Returns: The result from wolframscript as a string, or an error message.
No output schema documented. Tool returns unstructured string; LLM cannot plan downstream operations or extract structured data.
Error handling returns plain text strings instead of structured errors with recovery guidance. 'Error: wolframscript command not found' gives no actionable next step.
Tool description (127 chars) lacks critical context: when to use it, what Mathematica is, prerequisites (wolframscript installation), and timeout behavior. Does not explain what 'expression' means or what formats are valid.
Parameter 'expression' description is generic ('The Mathematica expression string to evaluate'). No guidance on syntax, valid formats, or examples of what constitutes a valid expression.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 45 | 2026-07-28+ | v2 |
No security considerations documented. Tool executes arbitrary code via wolframscript; no mention of sandboxing, input validation, or injection risks. Timeout is present (30s) but not documented in description.