Sec-Input-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 input sanitization that strips angle brackets but does not do context-aware output encoding.

{
  "metrics": {
    "correct": {
      "checks": [
        {
          "contains": [
            "context"
          ]
        },
        {
          "contains": [
            "encoding"
          ]
        },
        {
          "contains": [
            "sanitize"
          ]
        },
        {
          "contains": [
            "XSS"
          ]
        },
        {
          "contains": [
            "output"
          ]
        }
      ]
    },
    "hidden": {
      "checks": [
        {
          "contains": [
            "HTML"
          ]
        },
        {
          "contains": [
            "attribute"
          ]
        },
        {
          "contains": [
            "JavaScript"
          ]
        },
        {
          "contains": [
            "URL"
          ]
        },
        {
          "contains": [
            "escape"
          ]
        }
      ]
    }
  }
}