Separate reported-problem polarity from free-text tool details

When an LLM tool records a possible problem, carry an explicit typed problem flag alongside its descriptive text. Do not infer escalation from a nonempty description: an answer denying a problem can still be returned as a nonempty phrase. Let the model classify the caller's meaning, validate the structured arguments, and persist no problem detail when the flag is false. Test both a denial and a real problem through the handler. This removes string-truthiness errors; it does not establish that the model's semantic classification is always correct.

Applies when: voice tool results drive human-review or escalation decisions from nullable free text. Keywords: voice LLM tools Pydantic boolean issue escalation denial polarity validation