Single tool 'query' has a well-structured input schema with proper types, enums, and constraints (start/max_results bounds, sortBy/sortOrder enums). Description is adequate (89 chars) but generic, lacks WHEN to use it or what the output structure contains. Parameter descriptions are present and specific (e.g., 'arXiv search expression', 'Comma-delimited arXiv IDs'). However, output schema is not documented in the tool definition, LLMs cannot see what fields the response contains. Error handling is present in code (validation, bounds checking) but not surfaced in tool metadata. No tool annotations (readOnlyHint, etc.). Naming is clear and verb-forward ('query').

• Add explicit output schema documentation to the tool definition. Document the ParsedEntry structure (id, title, summary, published, updated, authors[], categories[], primary_category, links[], comment, journal_ref, doi) and the feed wrapper (title, id, updated, totalResults, startIndex).
• Expand tool description to 100-150 chars: 'Query arXiv articles by search expression or ID list. Returns structured metadata including title, authors, categories, and links. Use for literature discovery, citation tracking, or paper metadata retrieval.'
• Add readOnlyHint annotation to the tool definition to signal that this tool is safe to call repeatedly without side effects.
• Document error recovery in the tool description: 'If search returns no results, try broadening the search_query or check arXiv ID format (e.g., 2301.01234 or hep-th/9901001v1).'
• Add pagination guidance: 'Use start and max_results to iterate through large result sets. The response includes totalResults; paginate by incrementing start by max_results until start >= totalResults.'