This server has critical gaps across naming, descriptions, and parameter documentation. Tool names lack action verbs (e.g., 'articles', 'article', 'collections' instead of 'search_articles', 'get_article', 'search_collections'). Descriptions are extremely terse (5-20 chars), providing no context for LLM selection or usage guidance. Parameters lack descriptions entirely, 'item_type', 'order', 'page', 'page_size' have no explanations of valid values, ranges, or constraints. Output schemas are completely undocumented. Error handling is minimal (generic 'Figshare: not found' messages). The server reads from a public API with no authentication, but lacks pagination guidance, result limits, and field documentation. No tool annotations (readOnlyHint, etc.) despite all tools being read-only.

• Rename tools with action verbs: 'search_articles', 'get_article', 'get_article_files', 'search_collections', 'get_collection', 'list_institutions'. This makes intent clear from the name alone.
• Expand all descriptions to 50 - 150 characters. Example: 'Search Figshare articles by keyword, type, or date. Returns paginated results with metadata. Use get_article for full details.' This tells LLMs what, when, and what to expect.
• Add descriptions to all parameters. Example for 'item_type': 'Filter by content type: 1=figure, 3=dataset, 5=fileset, 12=thesis. Omit to search all types.' For 'order': 'Sort results by: published_date (default), modified_date, views, downloads, shares, or cites.'
• Document output schemas in tool descriptions or a separate schema field. Example: 'Returns {items: [{id, title, doi, published_date, authors}], total_count, page}.' This enables chaining and field extraction.
• Add readOnlyHint: true to all tool definitions (via tool annotations) since all tools are read-only queries. This signals to agents that retries are safe and no state changes occur.