Two tools with minimal but present schemas and descriptions. Tool naming follows verb_noun pattern (add, calculate), which is good for LLM parsing. However, descriptions are very short (11-43 chars, below baseline of 194 chars), parameter descriptions are minimal, and output schemas are not documented. Error handling exists for divide-by-zero but is not systematic. No tool annotations (readOnlyHint, destructiveHint, idempotentHint) despite correct risk classification. Schemas use Zod validation at runtime but descriptions lack detail about constraints, expected return formats, and when to use each tool.

• Expand tool descriptions to 100-200 characters, explaining WHAT the tool does, WHEN to use it vs. the other tool, and WHAT it returns. E.g., 'add: Simple addition of two numbers. Returns the sum as a string. Use when you need basic addition without choosing an operation. calculate: Performs arithmetic (add, subtract, multiply, divide). Returns the result as a string. Use when the operation is dynamic or you need subtract/multiply/divide.'
• Add parameter descriptions for 'operation', 'a', and 'b'. E.g., operation: 'One of: add, subtract, multiply, divide. Division by zero returns an error message instead of throwing.'; a: 'First operand (number)'; b: 'Second operand (number). For divide, cannot be zero.'
• Document output schema in tool metadata. The response structure { content: [{ type: 'text', text: <result as string> }] } should be declared so LLMs know what to expect.
• Add tool annotations to the server.tool() registration to declare readOnlyHint: true, since both tools are read-only and side-effect-free.
• Improve error handling for divide-by-zero and invalid operations. Return a structured error response with error classification (retryable: false, user_fixable: true) and actionable guidance. E.g., { error: 'invalid_operation', message: 'Operation must be one of: add, subtract, multiply, divide. Got: invalid', retryable: false }.
• Document numeric bounds for 'a' and 'b' parameters if there are known limits (e.g., max/min values, precision).