skill-2-agentic-concern-extraction

To extract structured concerns from agentic review outputs, run the agentic concern extractor on a paper's result directory. It integrates major/minor concerns, verdict drivers, and provenance from summary.yaml, review.md, adversarial_brief.md, gates.md, and scorecard.md into a single report.

An agentic concern sheet is a schema-compliant report that normalizes severities, deduplicates concerns, and records source details for traceability. You need one when building concern-alignment data for a paper to ensure consistent representation of concerns across reviews.

Normalize severities and deduplicate concerns by running the extractor across summary.yaml, review.md, adversarial_brief.md, gates.md, and scorecard.md. It consolidates duplicate concerns across sources and produces a schema-compliant AgenticConcernSheet with full provenance for downstream analysis.

The agentic concern extractor applies to single paper, single method, and single version result directories. It is designed to integrate decisive drivers, major/minor concerns, and decision drivers into a single report for one specific review scope.

Generating an AgenticConcernSheet requires agentic review outputs in YAML and Markdown formats, specifically summary.yaml, review.md, adversarial_brief.md, gates.md, and scorecard.md. These files provide the decisive drivers, concerns, and origins extracted into the final report.