This server has 17 tools with generally well-defined schemas and descriptions, but several critical gaps limit production readiness. Naming is clear and action-oriented (search_, get_, add_, remove_). All tools have non-empty descriptions (ranging 30-150 chars). Input schemas are present and use proper JSON Schema with types. However: (1) descriptions are somewhat generic and lack WHEN/WHY guidance for LLM tool selection; (2) some parameter descriptions are minimal (e.g., 'The ID of the product to add' for product_id); (3) no documented output schemas, LLMs cannot infer what fields to expect from responses; (4) no error handling guidance or recovery hints; (5) no tool annotations (readOnlyHint/destructiveHint); (6) password parameter exposed in initialization (see security issue below). The server is serviceable for domain-specific grocery shopping automation, but falls short of production-grade agent tooling standards. Average per-tool score: 58/100.

• Document output/response schemas for every tool. Example for search_products: 'Returns { products: [ { product_id: number, name: string, price: number, availability: boolean } ], total_count: number, next_cursor: string | null }'. Include this in tool registration or code comments.
• Expand descriptions to 50-200 characters and include decision guidance. Example: 'Search for products by name. Use this instead of get_account_data when you only need to find specific items. Returns product IDs for use with add_to_cart.'
• Add tool annotations: destructiveHint for add_to_cart, remove_from_cart; readOnlyHint for all get_* tools. This lets agents reason about retry safety and side effects.
• Wrap tool handlers in try-catch blocks with actionable error messages. Example: catch RohlikAPIError and return { error: 'Product not found. Try search_products() with a partial name or different keywords.' }
• Add min/max bounds to numeric parameters in descriptions. Example: 'limit (1-50, default 20): Maximum number of results to return. Larger values slow responses.'
• Document pagination support: if search_products returns 50 results, accept offset/limit params and return total_count so LLMs can iterate over large result sets without overflowing context.
• Clarify tool overlap in descriptions. Example: 'get_account_data returns everything including cart and orders (heavy payload). get_cart_content returns only cart items (lightweight). get_order_history returns only delivered orders (medium payload). Choose based on what you need.'