skills/ai-eval-ci/_scores.json

{
  "version": "1.0.0",
  "skillHash": "sha256:208bc64cea25b740aedab1d382d3b85da883f29f78f6ceec3585aeaa62b07688",
  "scoredAt": "2026-05-13T13:41:54.056Z",
  "backend": "ollama",
  "model": "gpt-oss:20b",
  "quality": {
    "score": 93,
    "dimensions": {
      "clarity": "PASS",
      "completeness": "WEAK",
      "conciseness": "PASS",
      "actionability": "PASS",
      "crossPlatform": "PASS",
      "examples": "PASS"
    },
    "issues": [
      {
        "severity": "MEDIUM",
        "category": "completeness",
        "detail": "The skill does not cover error handling for missing API keys, network failures, or provider errors."
      }
    ]
  },
  "security": {
    "verdict": "SAFE",
    "issues": []
  },
  "impact": {
    "multiplier": 9.3,
    "baselineAvg": 10,
    "treatmentAvg": 93,
    "scenarios": [
      {
        "name": "add-ci-quality-gates",
        "baseline": 0,
        "treatment": 95,
        "rationale": "Response B fully satisfies the rubric with a Promptfoo config, GitHub Actions workflow, and PR comment logic, while Response A provides no content."
      },
      {
        "name": "compare-models-eval",
        "baseline": 20,
        "treatment": 90,
        "rationale": "Response B fully meets the rubric by configuring Promptfoo with the required providers, running the eval suite, producing a comparison table with scores, latency, and cost, and recommending a model based on score-to-cost ratio, whereas Response A only provides a custom script without Promptfoo, lacking per-test metrics, cost, and recommendation."
      }
    ]
  }
}