Loading ToolBench
Native open-source terminal AI coding agent for Xiaomi MiMo
Two tools with basic schemas and descriptions, but significant gaps in parameter documentation and output schema clarity. Both tools have action-verb names (glob, grep) and non-empty descriptions (88 and 96 chars respectively, within baseline range). However, parameter descriptions are minimal or absent. The 'limit' parameter lacks description in both tools. Output schemas are not documented, LLMs cannot infer what fields are returned or how to chain results. Error handling guidance is absent. No tool annotations (readOnlyHint present in metadata but not in schema). Schemas are present and properly typed, but incomplete for production use.
List files in the workspace matching a glob pattern (e.g. **/*.rs). Respects .gitignore.
Search file contents in the workspace with a regex. Returns file:line:matched-text. Respects .gitignore.
Output schemas not documented. LLMs cannot infer what fields glob() and grep() return, preventing downstream tool chaining and forcing agents to guess at response structure.
Parameter 'limit' lacks description in both tools. LLMs cannot determine valid range, default behavior, or when to use it. Baseline: 100% of A+ tools have descriptions for every parameter.
No error handling guidance. Tools do not document what happens on invalid patterns, permission errors, or timeouts. LLMs receive raw errors with no recovery path.
Tool annotations missing from schema. Risk metadata (READ_ONLY) is in file comments but not in JSON Schema as readOnlyHint. Agents cannot parse this from code.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 46 | <=2025-11-25 | v2 |
No pagination or result limiting guidance in descriptions. If glob or grep return large result sets, LLMs have no signal to paginate or cap results, risking context window exhaustion.