Sec-Dependency-Risk-Classifier

The model receives the prompt (and optional system message). The run uses scorer rubric_json_metrics with the JSON configuration below. Pass/fail and partial credit are determined entirely by that scorer against the model output; no human grading.

Return JSON only with keys verdict, risk, mitigation. Review dependency risk classification that ignores maintainer activity, install scripts, provenance, and known advisories.

{
  "metrics": {
    "correct": {
      "checks": [
        {
          "contains": [
            "dependency"
          ]
        },
        {
          "contains": [
            "maintainer"
          ]
        },
        {
          "contains": [
            "install script"
          ]
        },
        {
          "contains": [
            "provenance"
          ]
        },
        {
          "contains": [
            "advisory"
          ]
        }
      ]
    },
    "hidden": {
      "checks": [
        {
          "contains": [
            "license"
          ]
        },
        {
          "contains": [
            "typosquat"
          ]
        },
        {
          "contains": [
            "SBOM"
          ]
        },
        {
          "contains": [
            "lockfile"
          ]
        },
        {
          "contains": [
            "risk"
          ]
        }
      ]
    }
  }
}