Loading ToolBench
Gene Ontology browser. Wraps the keyless EBI QuickGO API: search GO terms by keyword, get a GO term's definition/aspect/synonyms, and list the GO annotations for a gene/protein (by UniProt accession). Keyless.
Three tools with complete input schemas and clear descriptions. All tools are read-only, keyless, and well-named with action verbs (search_, get_, gene_). Descriptions are concise (100-150 chars) and explain what each tool does. However, output schemas are NOT documented, LLMs cannot see what fields to expect from responses. Parameter descriptions are minimal (e.g., 'Max results' lacks range/default clarity). No error handling guidance. No tool annotations (readOnlyHint present but not declared). Missing composition guidance for multi-step workflows.
QuickGO (EBI) — list the Gene Ontology (GO) annotations for a gene/protein, identified by UniProt accession (e.g. "P04637"). Returns GO ids, names, aspect, evidence, and taxon. Keyless.
QuickGO (EBI) — get a single Gene Ontology (GO) term by id. Returns its name, aspect (biological_process|molecular_function|cellular_component), definition, synonyms, and obsolete flag. Keyless.
QuickGO (EBI) — search Gene Ontology (GO) terms by keyword. Returns matching GO terms (id + name) for a free-text query like "apoptosis". Keyless.
Output schemas not documented. LLMs cannot see what fields responses contain (e.g., searchTerms returns {count, terms[{id, name}]} but this is invisible to the agent). Forces LLMs to guess field names and risks extraction errors.
Parameter descriptions lack actionable constraints. 'Max results (default 15, max 100)' is vague, does limit=0 work? What if omitted? Descriptions should state: 'Limit results to 1 - 100 (default 15). Omit to use default.'
No error handling guidance. If a GO term ID is invalid or a UniProt accession not found, the tool returns {error: '...'} but does not guide the LLM on recovery (e.g., 'Try search_terms() to find valid GO IDs'). Agents cannot self-correct.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 57 | 2026-07-28+ | v2 |
Tool annotations missing. All three tools are read-only (safe to retry), but readOnlyHint is not declared in the tool definition. LLMs cannot infer safety without explicit annotation.
No composition guidance. Tools are independent; no hints on typical workflows (e.g., 'search_terms() → get_term() → gene_annotations()'). Agents must discover patterns through trial.