Three tools with basic definitions and schemas, but significant gaps in LLM-optimization and parameter documentation. All tools have descriptions (10-50 chars baseline), but descriptions are generic and lack WHEN/HOW guidance per pattern:tool-description. Input schemas present with type info, but parameter descriptions are minimal (5-15 chars). No output schemas documented. No error handling guidance. Tool names are action-oriented (hello, add_numbers, search_items) but lack specificity, 'hello' and 'add_numbers' are trivial utilities, not production patterns. search_items has a reasonable description but lacks pagination guidance despite accepting max_results. No tool annotations (readOnlyHint, idempotentHint). Overall: template-quality, not production-ready.

• Expand parameter descriptions to 50-100 chars. For 'name', write: 'The full name of the person to greet (e.g., John Smith). Required.'
• Document output schemas explicitly. E.g., search_items returns: {results: [{id: int, name: string, score: float}], total: int, query: string}
• Add WHEN/HOW context to tool descriptions. E.g., 'Use when the user explicitly asks to greet someone by name. Returns a friendly greeting string.'
• Add pagination to search_items: accept offset/limit, return total_count and has_more flag. Cap results at 20 by default, document the limit in the description.
• Add error handling examples: 'If query is empty, return error: "Query cannot be empty. Provide a search term."' and document which errors are retryable.