skill-4-alignment-aggregate

Cross-paper pattern mining analyzes alignment datasets containing official concern sheets, agentic concern sheets, and concern-match graphs to surface systematic misses, severity biases, and judgment inversions. It applies a lint gate and derives severity from raw values to produce actionable improvement proposals.

Run an aggregate alignment skill after completing per-paper concern matching across many papers. It processes the resulting alignment datasets to generate a human-readable calibration report and a machine-readable proposals.yaml for governance and iteration.

Cross-paper pattern mining differs from single-paper review by analyzing alignment datasets across many papers to identify systematic rubric misses and severity biases. It outputs structured artifacts like a machine-readable proposals.yaml for governance rather than isolated per-paper adjustments.

Cross-paper pattern mining for rubric calibration is limited by its dependency on completed per-paper concern matching datasets containing official concern sheets, agentic concern sheets, and concern-match graphs. It cannot derive severity or output proposals without passing its enforced lint gate.