Loading ToolBench
Europe PubMed Central — biomedical literature search and full text
Server defines 5 tools with complete input schemas and basic descriptions. All tools are read-only and follow verb_noun naming (search, get_article, abstract, references, citations). However, descriptions are terse (10-80 chars, well below the 194-char baseline), lack WHEN/WHY context, and parameter descriptions are minimal or absent. The 'source' parameter appears in 4 tools but has no description explaining valid values (MED, PMC, PPR, CTX). Output schemas are not documented, LLMs cannot predict response structure for chaining. Error handling is minimal (404, generic HTTP errors). No tool annotations (readOnlyHint, idempotentHint) despite all being read-only. Composition is sound (each tool does one thing), but descriptions need expansion to meet production standards.
Just the title + abstract for one article (faster than get_article).
List of articles citing one article.
Full record for one article by (source, id). Source: MED (Medline), PMC, PPR (preprints), CTX (clinicaltrials).
List of references cited by one article.
Europe PMC search (lucene-style query). Returns metadata for matching articles.
Parameter descriptions missing or incomplete. 'source' parameter (used in 4 tools) has no description; LLMs cannot infer valid values (MED, PMC, PPR, CTX). 'id' parameter lacks context (PMID vs PMCID vs other formats). Violates pattern:tool-description baseline (100% of A+ tools have param descriptions).
Tool descriptions are terse (10 - 80 chars vs 194-char baseline). Lack WHEN/WHY context. 'abstract' description (47 chars) does not explain when to prefer it over get_article. 'references' and 'citations' descriptions (both ~50 chars) do not clarify the distinction or use cases.
Output schemas not documented. LLMs cannot predict response structure (fields, types, pagination). Violates pattern:tool baseline (100% of A+ tools have documented return types). Breaks tool chaining, agent cannot know what IDs to extract for follow-up calls.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 53 | 2026-07-28+ | v2 |
No tool annotations despite all tools being read-only. Missing readOnlyHint annotation would signal to agents that these tools are safe to call speculatively without side effects. Violates current spec (2026-07-28) which rewards tool annotations.
Error handling is minimal. Generic HTTP errors (404, 500) with no recovery guidance. Does not follow pattern:recovery-guide (errors should tell LLM what to do next). Example: 'Europe PMC: 404' gives agent no actionable path forward.