This server has 27 tools with basic descriptions and partial input schemas. Most tools lack comprehensive parameter documentation and output schema definitions. Naming is generally clear (verb_noun pattern), but descriptions are often generic (10-50 chars) and do not explain WHEN to use each tool or what it returns. No error handling guidance, no output schemas documented, and no tool annotations (readOnlyHint/destructiveHint). The instructions.py file shows good conversational guidance but does not translate to tool-level metadata. Many tools are READ_ONLY or WRITE, but this risk classification is not exposed in the tool definitions themselves.

• Add output schemas to all tools. Document the structure of returned objects (e.g., restaurant list returns [{id, name, rating, delivery_time, cuisine}]). This enables LLMs to chain calls correctly.
• Expand descriptions to 100-200 chars and include WHEN/WHY guidance. E.g., 'Retrieve current food cart contents before checkout to confirm items and total. Returns list of items with prices and quantities.'
• Add tool annotations: mark read-only tools with readOnlyHint=true, destructive operations (place_food_order, checkout, book_table) with destructiveHint=true, and idempotent tools with idempotentHint=true.
• Document error cases and recovery paths. E.g., 'If restaurant_id is invalid, returns error 404 with message. Try search_restaurants() to find valid restaurant IDs.'
• Add parameter descriptions for all tools, including format constraints. E.g., 'area: Delivery location (e.g., 'Rajkot', 'Mumbai'). Required for search.' and 'date: Booking date in YYYY-MM-DD format (e.g., 2025-01-15).'
• Implement pagination for list tools (get_food_orders, get_orders, search_restaurants). Add limit and offset/cursor parameters and return total_count to prevent context window exhaustion.

Inferred effective spec: <=2025-11-25.

• Relies on Logging (deprecated) - log to stderr or use OpenTelemetry