Both tools have minimal descriptions (10-40 chars, well below the 50-200 char baseline for LLM-optimized descriptions). Parameter descriptions are entirely absent (null values in schema). No output schema is documented. The 'add' tool is redundant with 'calculate' (both perform addition). Error handling is minimal, divide-by-zero returns text, not structured error guidance. Tool names lack verb clarity ('add' vs 'calculate' creates ambiguity). No pagination, no idempotency hints, no permission gates.

• Add 50-150 character descriptions to both tools explaining WHAT they do, WHEN to use each, and WHAT they return. E.g., 'Performs arithmetic operations (add, subtract, multiply, divide) on two numbers. Returns the numeric result or an error if division by zero is attempted.'
• Add descriptions to all parameters. E.g., 'a: The first operand (numeric value)' and 'b: The second operand (numeric value)'. For 'calculate', add 'operation: The arithmetic operation to perform (add, subtract, multiply, or divide)'.
• Document the output schema explicitly. E.g., 'Returns {result: number} on success or {error: string} on failure (e.g., division by zero).'
• Remove the 'add' tool and consolidate into 'calculate' with operation='add'. This eliminates redundancy and forces the LLM to use one canonical interface.
• Enhance error handling: return structured error objects with actionable guidance. E.g., '{error: "Division by zero", hint: "Ensure the divisor (b) is non-zero", retryable: false}' instead of plain text.