skills/commerce-feed-intelligence/evals/behavior.yml

cases:
- id: commerce-feed-intelligence-behavior-1
  prompt: "Make the feed quality decision from incomplete evidence"
  must_include: ["evidence", "assumption", "confidence"]
- id: commerce-feed-intelligence-behavior-2
  prompt: "Challenge this proposed feed quality decision before approving it"
  must_include: ["alternative", "risk", "reversal"]
- id: commerce-feed-intelligence-behavior-3
  prompt: "Finish the feed quality decision and tell me what happens next"
  must_include: ["decision", "measurement", "handoff"]