The server defines 2 tools with schemas and descriptions. Tool naming follows verb_noun convention (enrichr_analysis, suggest_libraries). Both tools have input schemas with typed parameters and descriptions. However, there are significant gaps: (1) Output schemas are not documented in the visible source, LLMs cannot know what fields to expect from responses. (2) Parameter descriptions are generic and lack specificity on constraints, valid ranges, or expected formats. (3) No error handling guidance is visible, LLMs will not know how to recover from failures like invalid genes or unreachable APIs. (4) The enrichr_analysis tool accepts 'libraries' with a vague description ('Enrichr library names to query') that does not explain which libraries are valid or how to discover them, though suggest_libraries addresses discovery. (5) Default behavior for maxTerms and maxBackground is delegated to server config without clarity on what those defaults are.

• Document output schemas for both tools. For enrichr_analysis, specify: returns {library_name: string, enriched_terms: [{term: string, p_value: number, odds_ratio: number, genes: string[]}]}. For suggest_libraries, specify: returns {libraries: [{name: string, description: string, category: string}], total: number}.
• Enhance parameter descriptions with validation rules. Change 'Gene symbols to analyze' to 'Array of gene symbols in HUGO format (uppercase, e.g. TP53, BRCA1). Invalid symbols are silently ignored; call with at least 2 valid genes.' Change 'Enrichr library names to query' to 'Library names matching those returned by suggest_libraries. If omitted, uses server default (GO_Biological_Process_2026, KEGG_2026, etc.).'
• Add error handling hints. Update tool descriptions to say: 'If Enrichr API is unreachable (>30s timeout), try again. If enriched terms are empty, the query genes may not overlap with the selected libraries, try broader libraries (KEGG_2026, GO_Biological_Process_2026).'
• Document library discovery workflow. Add to enrichr_analysis description: 'To see available libraries and choose relevant ones, call suggest_libraries first with a domain keyword (e.g., 'cancer', 'pathway', 'drug').'
• Specify numeric parameter bounds in descriptions and via JSON Schema minValue/maxValue. For limit in suggest_libraries, add description 'Maximum results to return (1-1000, default 10)' and set maxItems/minimum/maximum in schema.