JSON Contract (schema_version: "1.0")

When executing typesafe-eval ... -f json, structured data is emitted strictly to stdout. Diagnostic warnings, progress spinners, and file-level errors are sent to stderr.

---

JSON Schema Structure

[
  {
    "schema_version": "1.0",
    "filepath": "docs/architecture.md",
    "filename": "architecture.md",
    "preset_name": "tech-spec",
    "passed_thresholds": false,
    "composite_score": 0.52,
    "scores": {
      "clarity": {
        "score": 0.72,
        "max_score": 1.0,
        "normalized_score": 0.72,
        "confidence": 0.89,
        "probabilities": {
          "1": 0.05,
          "2": 0.23,
          "3": 0.72
        },
        "near_threshold": false
      }
    },
    "nouls": {
      "has_test_plan": {
        "probability": 0.32,
        "overridden_by": null,
        "near_threshold": true
      }
    },
    "choices": {
      "readiness": {
        "choice": "needs_revision",
        "confidence": 0.82,
        "probabilities": {
          "ready": 0.12,
          "needs_revision": 0.82,
          "blocked": 0.06
        }
      }
    },
    "violations": [
      "has_test_plan: probability 0.32 below threshold 0.50",
      "Credential Exposure: [REDACTED_AWS_KEY] is a known aws_key credential"
    ],
    "warnings": [
      "clarity: score 0.58 below threshold 0.60 (warning)"
    ],
    "email_evaluations": [
      {
        "placeholder": "[EMAIL_1]",
        "outcome": "personal",
        "probability": 0.95,
        "decided_by": "free_mail",
        "near_threshold": false
      }
    ],
    "secret_evaluations": [
      {
        "placeholder": "[REDACTED_AWS_KEY]",
        "outcome": "secret",
        "probability": null,
        "decided_by": "rule",
        "near_threshold": false
      }
    ]
  }
]

---

Field Reference

• schema_version: Version identifier for forward/backward compatibility ("1.0").
• filepath: Filepath on disk or synthetic identifier.
• filename: Evaluated file basename.
• preset_name: Name of the applied evaluation preset.
• passed_thresholds: Boolean status flag. true only if all gate thresholds are met and zero safety violations occurred.
• composite_score: Weighted aggregate score across evaluated score/noul questions (\(0.0 \dots 1.0\)).
• scores: Numerical dimensions mapped to ScoreResult objects (including score, confidence, probabilities, and near_threshold).
• nouls: Binary decision dimensions mapped to NoulResult objects (including probability and near_threshold).
• choices: Categorical dimensions mapped to ChoiceResult objects (including choice, confidence, and probabilities).
• violations: Array of actionable failure strings causing non-zero exit code.
• warnings: Non-blocking observations, advisory notes, and threshold warnings.
• email_evaluations: Records each masked email and whether it was decided as personal or role.
• secret_evaluations: Records each detected secret token without exposing raw values.