The server defines 5 tools with complete input schemas and descriptions. All tools follow verb_noun naming convention (list_, get_, search_). However, descriptions are minimal (10-64 chars, below the 194-char production baseline), and output schemas are not documented. Parameter descriptions are present but terse. No error recovery guidance beyond validation messages. The tools are read-only catalog queries with low composition complexity, appropriate for the domain, but descriptions lack context for LLM selection (when to use search_perfumes vs list_all_perfumes, etc.).

• Document output schemas for all tools in the inputSchema or as schema comments. Example for get_perfume_by_id: 'Returns {success: bool, perfume: {id: string, name: string, brand: string, category: string, gender: string, description: string, price: number}}.'
• Expand descriptions to 80-150 characters with WHEN context. Example: 'Search for perfumes by name or brand fragment. Use this when the user mentions a partial name or brand; use list_all_perfumes() for browsing and get_perfume_by_id() when you have an exact ID.'
• Add error recovery hints to error messages. When perfume not found, return: 'Perfume with ID "<id>" not found. Try search_perfumes(query="<partial_name>") to find similar perfumes.'
• Implement pagination for list_all_perfumes and search_perfumes. Add optional 'limit' (default 20, max 100) and 'offset' (default 0) parameters. Return {total_count, returned_count, perfumes[...], has_more}.
• Document the perfume object schema inline or as a type definition. Specify required fields (id, name, brand) vs optional (description, price, rating).