PetstoreAPI has 10 tools with basic naming and descriptions, but significant gaps in schema completeness, parameter documentation, and error handling. Tool names follow verb_noun convention (good), but descriptions are generic and lack actionable context. Most parameters lack type constraints (enums, ranges). Output schemas are undocumented. Error handling returns raw API responses without recovery guidance. The query_petstore tool is a meta-tool that delegates to other tools, violating single-responsibility principle. No tool annotations (readOnlyHint, destructiveHint) despite clear read/write distinctions.

• Remove query_petstore entirely. It adds no value, agents should call specific tools directly. If natural-language routing is needed, implement it in the agent layer, not as a tool.
• Add enum constraints to all status/choice parameters. E.g., status: {type: 'string', enum: ['available', 'pending', 'sold']} for find_pets_by_status. This prevents hallucinated values.
• Document output schemas for every tool. E.g., 'Returns: {pet_id: int, name: string, status: string, tags: [string]}'. Include this in the tool description or as a separate schema field.
• Enhance error messages with recovery guidance. E.g., 'Pet ID 999 not found. Try find_pets_by_status("available") to list available pets.' instead of raw 404.
• Add tool annotations: @mcp.tool(..., readOnlyHint=True) for get_* and find_* tools; @mcp.tool(..., destructiveHint=True) for place_order and create_user. This signals to agents which operations are safe to retry.