Loading ToolBench
SkillRoute MCP stdio server for routing requests to skills and searching a skill catalog
Three tools with complete input schemas and descriptions, but descriptions are generic and lack LLM-optimized guidance. All tools are read-only (good for safety), but descriptions don't explain WHEN to use each tool vs. alternatives, dependencies between them, or expected output structure. Parameter descriptions are present but minimal. No output schemas documented. Error handling guidance absent.
Fetch metadata, relationships, excerpts, and source references for one skill.
Return ranked skills, reasons, confidence, evidence snippets, suggested order, and clarification questions.
Search indexed skills using hybrid metadata, lexical, and selected backend retrieval signals.
Descriptions lack LLM-optimized guidance on WHEN to use each tool. 'Return ranked skills...' tells WHAT but not WHEN (vs. search) or dependencies. Descriptions should be 50-200 chars with clear selection criteria.
No output schemas documented. LLMs cannot plan downstream calls or extract fields without knowing response structure. Responses should include skill_id, confidence, evidence, relationships for chaining.
Parameter descriptions are minimal (e.g., 'User request or task to route to skills' for 'request'). Should include format hints, constraints, and examples of valid input to guide LLM usage.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 57 | 2026-07-28+ | v2 |
No error handling guidance. If catalog path is invalid, backend unavailable, or skill not found, LLM receives no recovery hints. Should return actionable errors like 'Catalog not found at <path>. Ensure SKILLROUTE_CATALOG is set.'
Tool naming uses dots (skillroute.route) instead of underscores (skillroute_route). Dots are non-standard in MCP tool names and may confuse LLM parsing. Use verb_noun convention with underscores.