skills/agent-memory/_scores.json

{
  "version": "1.0.0",
  "skillHash": "sha256:41cdf645f2420451761f85f00a17f4fda11dacadf61f316bacc50a48e87b3257",
  "scoredAt": "2026-05-13T12:12:32.704Z",
  "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": "Skill does not cover error handling or edge cases such as missing API keys or network failures."
      }
    ]
  },
  "security": {
    "verdict": "SAFE",
    "issues": []
  },
  "impact": {
    "multiplier": 1.81,
    "baselineAvg": 31,
    "treatmentAvg": 56,
    "scenarios": [
      {
        "name": "setup-file-based-memory",
        "baseline": 12,
        "treatment": 78,
        "rationale": "Response A covers most rubric items with a FileMemory class and hooks, missing only periodic consolidation and exact method name, while Response B lacks the required directory structure and class definition."
      },
      {
        "name": "semantic-search-sqlite",
        "baseline": 50,
        "treatment": 30,
        "rationale": "Response B provides a more complete implementation (database, embeddings, search) than the truncated Response A, but neither meets all rubric requirements."
      },
      {
        "name": "chroma-production-memory",
        "baseline": 30,
        "treatment": 60,
        "rationale": "Response A provides a more complete implementation with store, recall, deduplication, and category filtering, whereas Response B is incomplete and lacks the required methods and logic."
      }
    ]
  }
}