skills/ai-pentesting/_scores.json

{
  "version": "1.0.0",
  "skillHash": "sha256:8ac7e9fb1ec2e93d29b2d1faaad47430909c83c702ed5546ea807079bb4fa38c",
  "scoredAt": "2026-05-13T13:43:52.799Z",
  "backend": "ollama",
  "model": "gpt-oss:20b",
  "quality": {
    "score": 77,
    "dimensions": {
      "clarity": "PASS",
      "completeness": "WEAK",
      "conciseness": "PASS",
      "actionability": "PASS",
      "crossPlatform": "FAIL",
      "examples": "PASS"
    },
    "issues": [
      {
        "severity": "MEDIUM",
        "category": "completeness",
        "detail": "Instructions do not cover error scenarios or edge cases."
      },
      {
        "severity": "HIGH",
        "category": "crossPlatform",
        "detail": "Skill is vendor-locked to Shannon and OpenAI, limiting cross-platform compatibility."
      }
    ]
  },
  "security": {
    "verdict": "SUSPICIOUS",
    "issues": [
      {
        "severity": "MEDIUM",
        "category": "checksum",
        "detail": "Git clone of Shannon repository lacks checksum verification."
      }
    ]
  },
  "impact": {
    "multiplier": 8.5,
    "baselineAvg": 0,
    "treatmentAvg": 85,
    "scenarios": [
      {
        "name": "run-shannon-pentest",
        "baseline": 0,
        "treatment": 70,
        "rationale": "Response B covers most rubric items but fails to use REPO=./ and omits the required ./shannon wait command, whereas Response A is empty."
      },
      {
        "name": "ci-cd-pentest-integration",
        "baseline": 0,
        "treatment": 100,
        "rationale": "Response A fully satisfies every rubric requirement, while Response B omits the required Shannon commands and steps."
      }
    ]
  }
}