Two tools with basic schemas and descriptions, but significant gaps in parameter documentation, error handling, and output structure. search_papers has a clear description and proper input schema with types. extract_info lacks output schema documentation. Neither tool documents error cases or recovery paths. Parameter descriptions are minimal (e.g., 'The topic to search for' lacks format/constraint guidance). No tool annotations (readOnlyHint, destructiveHint, idempotentHint). Output is unstructured (plain strings/lists) rather than typed objects. Missing pagination guidance for search_papers despite max_results parameter. No validation or error recovery guidance in descriptions.

• Add output schema documentation to both tools. search_papers should document it returns {paper_ids: string[], count: int}. extract_info should return {title: string, authors: string[], summary: string, pdf_url: string, published: string} or error object.
• Enhance parameter descriptions with format/constraint guidance: topic (1-200 chars, alphanumeric + spaces), max_results (1-100, default 5), paper_id (arXiv format: YYMM.NNNNN or YYMM.NNNNNvN).
• Add error recovery guidance to descriptions. E.g., extract_info: 'Returns error if paper not found. Call search_papers(topic=...) first to populate the papers directory.'
• Implement structured error responses. Instead of plain strings, return {success: bool, data?: object, error?: {code: string, message: string, recovery_hint: string}}.
• Add tool annotations: search_papers should have destructiveHint=true (writes to disk); extract_info should have readOnlyHint=true. Declare idempotency where applicable.