Loading ToolBench
MCP server that fetches, dedupes, and scores FemTech industry & research items.
Two tools with basic schemas and descriptions, but significant gaps in parameter documentation and output schema clarity. radar_collect has an enum constraint on 'section' (good), but 'since' and 'limit' lack descriptions explaining format, range, or defaults. radar_sources has no parameters but minimal context. Neither tool documents output structure for LLM planning. Descriptions are present but generic (under 200 chars baseline). No error handling guidance, no recovery paths, no idempotency hints. Tool names follow verb_noun pattern (radar_* prefix is domain-specific but acceptable). Overall: functional but below production quality.
Fetches, deduplicates, and scores FemTech items from a specified section (industry, research, opportunities, discussions, or china) within a time window.
Returns the list of data sources available for each radar section (industry, research, opportunities, discussions, china).
Parameter descriptions missing or incomplete. 'since' lacks format guidance (ISO 8601 stated in description but not in param docs); 'limit' has no range constraints (default 15 stated in description, not in schema). LLMs cannot infer valid ranges or formats from names alone.
Output schema not documented. radar_collect returns { items: RadarItem[], warnings: string[] } but LLM has no visibility into RadarItem structure (fields, types, required keys). Prevents downstream tool chaining and forces LLM to guess field names.
No error handling or recovery guidance. Tools do not document failure modes (invalid section, network timeout, malformed date) or what the LLM should do next. A bare error response leaves the agent stuck.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 50 | 2026-07-28+ | v2 |
radar_sources description is generic ('Returns the list of data sources'). Does not explain when to call it, what structure it reveals, or how it relates to radar_collect. Discovery tools should guide multi-step planning.
No idempotency or side-effect hints. Both tools are read-only (correct Risk classification), but descriptions do not state this explicitly. LLMs benefit from knowing a tool is safe to retry without side effects.