The wisdom-mcp-local server defines 15 tools with complete input schemas and descriptions visible in the source code. All tools follow verb-noun naming convention (get_*, list_*, search_*, add_*, update_*, delete_*, toggle_*, generate_*). Descriptions are present for all tools (average ~80-120 chars) and for most parameters. However, descriptions lack depth and actionability, they state WHAT the tool does but often omit WHEN to use it, dependencies, or error recovery guidance. Parameter descriptions are generally present but some are generic ('The ID of the quote'). No output schemas are documented in the source, forcing LLMs to infer return structure. Error handling and recovery guidance are absent. No tool annotations (readOnlyHint, destructiveHint, idempotentHint) are visible. The server uses STDIO transport, which is not remotely accessible.

• Document output schemas for all tools in the tool definition or server documentation. Example: get_random_quote returns {id: string, text: string, author: string, source: string, category: string, theme: string, tags: string[], notes?: string, is_favorite?: boolean}. This enables LLMs to compose tools and extract fields with confidence.
• Add tool annotations to every tool definition: set readOnlyHint=true for all retrieval tools (get_*, list_*, search_*); set destructiveHint=true for delete_quote; set idempotentHint=true for operations that don't change on repeated calls (e.g. toggle_favorite called twice with same ID returns to same state, but side effects differ, clarify).
• Enhance tool descriptions with WHEN to use them. Example: 'Get a random quote from any author. Use this for general inspiration or when you want serendipity. If you need a quote on a specific topic, use search_quotes or get_random_quote_by_category instead.' This helps LLMs disambiguate similar tools.
• Add pagination support to search_quotes: include limit (default 20, max 100) and offset (default 0) parameters. Document: 'Returns up to limit results. Use offset to retrieve additional pages. Response includes total_count for pagination planning.'
• Convert allowed values in descriptions to enum constraints. For category, theme, and other controlled fields, use: {"type": "string", "enum": ["list", "of", "values"]} instead of listing examples in the description.
• Add error handling examples in the server documentation or tool descriptions. E.g. 'If quote_id does not exist, returns {error: "Quote not found. Try search_quotes() to find a quote first."}'. This guides LLM recovery.