Score

Use 2–10 levels. Level numbers start at zero. Write each level so it can be understood on its own.

{
  "impact": {
    "type": "score",
    "instructions": "Rate the reported impact.",
    "criteria": ["No task blocked", "A task is blocked but a workaround exists", "A task is blocked with no workaround"]
  }
}

For probabilities [0.05, 0.30, 0.65], the score is 0 × 0.05 + 1 × 0.30 + 2 × 0.65 = 1.60. It is a weighted mean — not a percentage.

The response includes the score, probabilities, a legend that maps level numbers to descriptions, and confidence. Confidence measures concentration around the modal level. It does not establish correctness.

Levels can be structured objects with examples. The response preserves them in the legend. To compare rubrics of different lengths, divide each score by number_of_levels - 1. Choose weights and action thresholds from evaluation data.