Sec-Url-Sanitizer

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 URL sanitization that allows javascript:, data:, backslash tricks, and encoded control characters.

{
  "metrics": {
    "correct": {
      "checks": [
        {
          "contains": [
            "javascript:"
          ]
        },
        {
          "contains": [
            "data:"
          ]
        },
        {
          "contains": [
            "URL"
          ]
        },
        {
          "contains": [
            "sanitize"
          ]
        },
        {
          "contains": [
            "scheme"
          ]
        }
      ]
    },
    "hidden": {
      "checks": [
        {
          "contains": [
            "canonicalize"
          ]
        },
        {
          "contains": [
            "decode"
          ]
        },
        {
          "contains": [
            "allowlist"
          ]
        },
        {
          "contains": [
            "control character"
          ]
        },
        {
          "contains": [
            "backslash"
          ]
        }
      ]
    }
  }
}