Sec-Abuse-Detection-Rate-Window

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. Login rate limiting uses only a per-IP one-minute counter, while attackers distribute attempts across many IPs against the same account.

{
  "metrics": {
    "correct": {
      "checks": [
        {
          "contains": [
            "rate limiting"
          ]
        },
        {
          "contains": [
            "per-IP"
          ]
        },
        {
          "contains": [
            "same account"
          ]
        }
      ]
    },
    "hidden": {
      "checks": [
        {
          "contains": [
            "account-based limit"
          ]
        },
        {
          "contains": [
            "sliding window"
          ]
        },
        {
          "contains": [
            "distributed attacks"
          ]
        }
      ]
    }
  }
}