Server has 27 tools with consistent naming (verb_noun pattern: search_*, get_*, extract_*, etc.) and documented schemas. All tools have descriptions (50-150 chars, within baseline 34-392 range). Input parameters include types and descriptions. However, output schemas are not explicitly documented in the source code, only inferred from tool names and descriptions. Error handling is minimal; no recovery guidance or actionable error messages visible. Parameters lack constraints (enums, ranges, format specs) that would guide LLM input validation. No tool annotations (readOnlyHint, destructiveHint, idempotentHint) despite all tools being READ_ONLY. Missing dependency hints and multi-step guidance.

• Document output schemas for all search and lookup tools. Specify: field names (e.g., 'id', 'title', 'authors', 'year'), types (string, number, array), and pagination structure (totalCount, nextCursor, or limit/skip). Example: search_publications returns {totalCount: number, data: [{id, title, authors: string[], year: number, doi: string}]}.
• Add enum constraints to parameters accepting a fixed set of values. For 'aggregationFunction', declare enum: ['sum', 'count', 'avg', 'min', 'max']. For 'sortBy', document valid field names or reference describe_schema. For 'system' in classify, enumerate supported classification systems.
• Add readOnlyHint: true to all 27 tool definitions. This signals to agents that these tools are safe to retry and have no side effects, enabling more aggressive caching and retry strategies.
• Enhance error handling in execute_dsl and fetch_search_pages. Document: (1) what 'large' means (e.g., >10,000 results), (2) what happens if confirmLargeFetch is false, (3) how to paginate if results exceed limits. Example: 'If results exceed 10,000, set confirmLargeFetch=true or use limit/skip for pagination.'
• Add discovery guidance in tool descriptions. Example: 'Call describe_schema first to see available fields for the entityType parameter.' For classify, add: 'Call describe_schema to list supported classification systems.' This prevents wasted calls and guides multi-step planning.

Inferred effective spec: <=2025-11-25.

• Relies on Logging (deprecated) - log to stderr or use OpenTelemetry