Loading ToolBench
Server has 4 tools with complete schemas and descriptions, but quality is uneven. All tools have input schemas with Zod validation and descriptions present. However, descriptions are minimal (10-80 chars, well below the 194-char baseline), lack context on WHEN to use tools, and omit recovery guidance. add_memory and delete_memory lack output schema documentation. search_memory has the strongest definition with threshold/limit constraints. No tool annotations (readOnlyHint/destructiveHint) despite clear risk levels. Error handling returns generic success/error flags without actionable recovery paths. Parameter descriptions are sparse, e.g., 'Session identifier (optional)' doesn't explain what a session is or how it affects behavior.
Stores a new memory entry for retrieval later
Deletes a specific memory by ID
An example tool that processes messages
Searches for memories based on semantic similarity to the query
Minimal parameter descriptions lack actionable context. 'Session identifier (optional)' does not explain what a session is, how it filters results, or when to provide it. LLMs cannot infer parameter semantics from bare names.
No output schema documentation. add_memory and delete_memory return {success, id/error, message} but this structure is not formally documented. LLMs cannot plan downstream calls without knowing response fields.
Error responses lack recovery guidance. Returning {success: false, error: 'Memory not found'} gives the agent no next step. Should suggest 'Try search_memory() to find existing memories' or list available options.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 49 | 2026-07-28+ | v2 |
example_tool is a placeholder with generic name and minimal description. 'An example tool that processes messages' does not explain what processing occurs, when to use it, or what it returns. Should be removed or replaced with a real tool.
No tool annotations despite clear risk levels. delete_memory is DESTRUCTIVE but lacks destructiveHint. search_memory is READ_ONLY but lacks readOnlyHint. Annotations help LLMs reason about safety and retry logic.