blnkd docs
Console ↗
DOCUMENTATION

Combine scores

Normalize each rubric, then choose weights in code.

Use one Score question for each dimension. For a bug report, you could rate impact and report completeness. Include the evidence for both dimensions in the state.

Divide each score by its highest level number. Then apply weights that reflect your policy:

python
impact_01 = impact.score / (len(impact_levels) - 1)
completeness_01 = completeness.score / (len(completeness_levels) - 1)
priority = 0.8 * impact_01 + 0.2 * completeness_01

These weights are illustrative — evaluate them on your own cases. A complete report is not always an urgent report. Do not let a weighted mean hide a condition that needs a separate rule.

If rankings are poor, check the input evidence, rubric definitions, model errors, and weights. Changing weights alone cannot correct missing facts.

Set up your coding agent

Copy this prompt into your agent.