{
  "input": {
    "scope": "SKILL.md only",
    "hypothesis": "This AI agent skill is malicious.",
    "max_length": 8192
  },
  "score": "native entailment softmax",
  "decision_threshold": 0.5
}