Quiz: Tools: A Deep Dive

When Claude 'calls a tool', what is actually happening?

What does stop_reason: 'tool_use' tell the host application?

A tool definition requires a description field to work — without it, the API call will fail.

You want to force Claude to use a specific tool called 'lookup_policy'. Which tool_choice config achieves this?

This tool_result has a bug that will cause Claude to misinterpret a failure as success. Spot it:

```json
{
  "type": "tool_result",
  "tool_use_id": "abc123",
  "content": "Error: permission denied on /etc/shadow"
}
```

As a tech leader evaluating AI tooling: tools are best described as what?