Critical gaps across all dimensions. Tools 1-10 (server.py) have descriptions and basic schemas, but tools 11-31 (toc_server.py) have empty input schemas with no visible parameter definitions. 27 of 31 tools lack actionable parameter descriptions. No output schemas documented. Error handling is minimal. Tool naming is inconsistent (member_query vs browse_menu). No tool annotations (readOnlyHint/destructiveHint). The codebase shows mock data and formatters but actual parameter validation and error recovery guidance are absent.

• For tools 11-31 (toc_server.py), add explicit input schemas with typed parameters and descriptions. Example: browse_menu should declare parameters like 'category' (enum: coffee|tea|food), 'limit' (int, 1-100, default 20), 'sort_by' (enum: popularity|price|rating).
• Expand tool descriptions to 50-200 characters following LLM-optimized format: 'WHAT it does, WHEN to use it, any prerequisites.' Example: 'Browse available coffee menu items filtered by category. Call this first to see available drinks before creating an order.'
• Document output schemas for all tools. Example for browse_menu: returns {items: [{id, name, price, category, description}], total_count, has_more}. This enables agents to extract IDs for downstream calls.
• Add tool annotations: mark read-only tools with readOnlyHint=true; mark write operations (create_order, claim_all_coupons) with destructiveHint=true and idempotentHint=false to guide agent retry logic.
• Implement error recovery guidance. When a lookup fails (e.g., member not found), return available alternatives or suggest next steps. Example: 'Member ID not found. Try member_query() with mobile number or open_id first.'