Two tools present with basic schemas and descriptions, but significant gaps in quality. Both tools have descriptions (10-50 chars, below the 194-char baseline), minimal parameter documentation, and no output schema documentation. No error handling guidance, no parameter constraints (enums), and no indication of idempotency or side effects. Tool naming follows verb_noun convention (add, get_capital_city) but descriptions are too terse to guide LLM selection. No evidence of composition patterns or chaining support.

• Expand tool descriptions to 100-200 chars. Example: 'add' → 'Adds two integers and returns their sum. Use this for arithmetic calculations. Returns an integer result.'
• Document output schemas explicitly. For 'add': 'Returns: {"result": integer}'. For 'get_capital_city': 'Returns: {"capital": string, "country": string}'.
• Add descriptions to all parameters. For 'add': 'a: First integer operand (required)', 'b: Second integer operand (required)'.
• Include error recovery hints in descriptions. E.g., 'get_capital_city: If country not found, returns error with list of available countries. Try search_countries() for partial matches.'
• Add tool annotations to the schema (readOnlyHint: true for both tools) to signal to LLMs that these are safe, non-destructive operations.