Three tools with significant definition gaps. search_and_summarize has a basic description but no parameter descriptions. send_message_endpoint lacks channel_id parameter visibility and has incomplete schema. get_messages_endpoint has better structure but all three tools lack error handling guidance, output schema documentation, and LLM-optimized descriptions. Tool names are verb-based but descriptions are generic (10-50 chars, well below 50-200 char baseline for A+ tools). No tool annotations (readOnlyHint/destructiveHint). Parameter descriptions missing entirely for most inputs.

• Add parameter descriptions to all inputs. E.g. 'query (string): Search terms for Tavily. Max 500 chars. E.g. "latest AI breakthroughs 2025".'
• Document output schemas explicitly. E.g. 'Returns {summary: string (max 2000 chars), sources: [string]}' for search_and_summarize.
• Expand tool descriptions to 100-150 chars with WHEN/WHY context. E.g. 'search_and_summarize: Fetch and summarize web results for a query. Use when you need current information not in training data. Returns a concise summary with source links.'
• Add error recovery guidance. E.g. 'If Tavily API fails, suggest: "Search service unavailable. Try a simpler query or check your API key."'
• Make send_message_endpoint schema explicit: add optional 'channel_id' parameter (defaults to DEFAULT_CHANNEL_ID if omitted) so LLM can see the option.