{
  "title": "Completed voice-agent conversation review rubric",
  "version": "draft-2026-10-26",
  "status": "Proposed template; no evaluation has been run and no grades are supplied.",
  "case_fields": ["internal_case_id", "provider_call_id", "agent_version", "rubric_version", "policy_reference", "speaker_labeled_transcript", "available_audio_reference", "tool_results", "timing_definition"],
  "criteria": [
    {"id": "task_understanding", "question": "Did the agent identify the requested task accurately?", "required_evidence": ["transcript", "task_context"]},
    {"id": "supported_claims", "question": "Do claims about completed actions agree with authoritative tool results?", "required_evidence": ["transcript", "tool_results"]},
    {"id": "next_step", "question": "Does the final response accurately describe the next available step?", "required_evidence": ["transcript", "policy_reference"]},
    {"id": "audio_review", "question": "Does the audio satisfy the team's documented audio criteria?", "required_evidence": ["audio", "audio_rubric"]}
  ],
  "criterion_outcomes": ["pass", "fail", "insufficient_evidence"],
  "automated_result_record": {"storage": "Separate from independent human review", "fields": ["judge_version", "criterion_id", "outcome", "evidence_turn_ids", "explanation"]},
  "independent_review_record": {"instructions": "Show the case without automated grades first. Leave rating and notes blank until reviewed.", "fields": ["case_id", "conversation_context", "exact_displayed_response", "rating", "reviewer_notes", "suggested_improvement"], "rating_options": ["Good", "Acceptable", "Not acceptable", "Bad"], "rating": null, "reviewer_notes": "", "suggested_improvement": ""},
  "release_decision": {"owner": "To be assigned", "policy": "Define consequential failures and disagreement resolution before running the review."}
}