Use when turning support tickets, user complaints, helpdesk transcripts, or failed assistant responses into reusable eval cases with expected behavior, assertions, and failure modes.
