Sec-Vulnerability-Scanner

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 a dependency vulnerability scanner that only checks direct dependencies and ignores lockfile transitive packages.

{
  "metrics": {
    "correct": {
      "checks": [
        {
          "contains": [
            "dependency"
          ]
        },
        {
          "contains": [
            "transitive"
          ]
        },
        {
          "contains": [
            "lockfile"
          ]
        },
        {
          "contains": [
            "CVE"
          ]
        },
        {
          "contains": [
            "vulnerability"
          ]
        }
      ]
    },
    "hidden": {
      "checks": [
        {
          "contains": [
            "SBOM"
          ]
        },
        {
          "contains": [
            "advisory"
          ]
        },
        {
          "contains": [
            "severity"
          ]
        },
        {
          "contains": [
            "patch"
          ]
        },
        {
          "contains": [
            "dev dependency"
          ]
        }
      ]
    }
  }
}