Server has 4 tools with basic descriptions and parameter schemas, but significant gaps in quality. All tools have descriptions (10-200 chars range), and input schemas are present with type definitions. However, descriptions lack actionable context for LLM selection, parameter descriptions are minimal, output schemas are not documented, and error handling guidance is absent. No tool annotations (readOnlyHint/destructiveHint) despite clear risk profiles. Naming is verb-forward but generic. Per-tool average: search_papers=48, get_paper_pdf=45, resolve_doi=52, search_arxiv=58.

• Document output schemas for all tools. Specify: search_papers returns array of {title, authors, year, abstract, doi, arxiv_id, citation_count, pdf_links}; get_paper_pdf returns {file_path, success, error_message}; etc. Include field types and whether fields are always present.
• Expand parameter descriptions to 50-100 chars with constraints. E.g. 'Search query: keywords, paper title, DOI (10.xxxx/xxx), or arXiv ID (1234.5678). Max 200 chars.' instead of 'Search query'.
• Add error recovery guidance to tool descriptions. E.g. 'If PDF unavailable, try resolve_doi() to find alternative open-access sources. If DOI invalid, use search_papers() to locate the paper first.'
• Add tool annotations: mark get_paper_pdf with destructiveHint=true (writes to disk); mark search_* with readOnlyHint=true. This enables agent to reason about side effects and plan accordingly.
• Clarify get_paper_pdf 'url' parameter: specify whether it accepts DOI URLs (https://doi.org/10.xxxx), landing pages (https://example.com/paper), or direct PDF links. Document fallback behavior if multiple params provided.

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

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