Scenario

You're building an LLM agent that can call tools like search, calendar lookup, and ticket creation. The model is generally good at choosing the right tool, but tool calls sometimes contain missing fields, wrong types, out of range values, or unsafe arguments copied from user input.

Question

Walk me through how you would handle parameter validation for tool calls in an LLM agent.