Loading ToolBench
Serve only the AI rules that match the file you're working on — over MCP, just in time.
Single tool 'get_rules' has a clear verb-noun name and acceptable description (108 chars, within 10-1024 range). However, the input schema is incomplete: 'task' parameter lacks a description entirely, and 'path' description is minimal (34 chars). The schema does not document output structure, pagination, or error cases. No parameter type validation (e.g., path format, task length constraints). The tool is read-only but lacks explicit idempotent/readOnly annotations. Overall definition quality is below baseline for production tools.
Fetch the ZeroContext rules that apply to a given file before editing it. Call with the file path you are about to modify.
Missing parameter description: 'task' parameter has no description. LLMs cannot infer whether it is required, optional, or how to format it.
No output schema documented. Tool returns text content but does not specify structure, fields, or format of the rules returned. LLMs cannot plan downstream actions.
No error handling guidance. Tool does not document what happens if path is invalid, file not found, or rules cannot be computed. No recovery hints for LLM.
Missing tool annotations. Tool is read-only but does not declare readOnlyHint=true. Agents cannot distinguish safe tools from destructive ones.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 54 | 2026-07-28+ | v2 |
Minimal parameter constraints. 'path' accepts any string; no validation for file existence, relative path format, or length limits. Invites invalid input from LLMs.