{
  "schema_version": 1,
  "total_cases": 5,
  "total_results": 5,
  "tested_cases": 5,
  "pass_rate": 1.0,
  "verdict_counts": {
    "pass": 5,
    "fail": 0,
    "inconclusive": 0,
    "untested": 0,
    "missing": 0
  },
  "failed_cases": [],
  "missing_result_case_ids": [],
  "unrecognized_result_case_ids": [],
  "surface_summary": {
    "Prompt-bearing agent instruction boundary": {
      "pass": 1,
      "fail": 0,
      "inconclusive": 0,
      "untested": 0,
      "missing": 0
    },
    "LLM-enabled agent security benchmark or survey surface": {
      "pass": 3,
      "fail": 0,
      "inconclusive": 0,
      "untested": 0,
      "missing": 0
    },
    "Agent command execution or command-classification boundary": {
      "pass": 1,
      "fail": 0,
      "inconclusive": 0,
      "untested": 0,
      "missing": 0
    }
  },
  "context_sufficiency": {
    "status": "sufficient",
    "covered_facts": [
      "case_ids",
      "verdict_counts",
      "failed_case_evidence",
      "surface_summary"
    ],
    "reason": "Summary is computed from generated eval cases and operator result rows."
  }
}