skills/ab-test-setup/_scores.json

{
  "version": "1.0.0",
  "skillHash": "sha256:a3ede63973cd0ddd7a1b812b1fe03d920316be8b1cfc81976789aa699321b642",
  "scoredAt": "2026-05-13T12:12:35.296Z",
  "backend": "ollama",
  "model": "gpt-oss:20b",
  "quality": {
    "score": 100,
    "dimensions": {
      "clarity": "PASS",
      "completeness": "PASS",
      "conciseness": "PASS",
      "actionability": "PASS",
      "crossPlatform": "PASS",
      "examples": "PASS"
    },
    "issues": []
  },
  "security": {
    "verdict": "SAFE",
    "issues": []
  },
  "impact": {
    "multiplier": 1.6,
    "baselineAvg": 53,
    "treatmentAvg": 85,
    "scenarios": [
      {
        "name": "landing-page-cta-color-test",
        "baseline": 45,
        "treatment": 100,
        "rationale": "Response A fully meets all rubric items, while Response B lacks hypothesis, sample size, metrics, and traffic split details."
      },
      {
        "name": "product-page-mvt-headline-image",
        "baseline": 35,
        "treatment": 55,
        "rationale": "Response B covers more rubric items (hypothesis, metrics, server‑side, checklist) but still misses MVT and the exact sample‑size calculation; Response A lacks hypothesis, metrics, and sample‑size, so it scores lower."
      },
      {
        "name": "analyze-checkout-form-test",
        "baseline": 80,
        "treatment": 100,
        "rationale": "Response A fully covers all rubric items, including pre‑launch checklist references, while Response B omits explicit mention of the pre‑launch checklist and guardrail significance."
      }
    ]
  }
}