The Evidence Self-loop skill acts as a quality gate for research-intensive writing pipelines, ensuring that every claim is backed by rigorous data. By auditing intermediate artifacts like evidence bindings, subsection briefs, and paper notes, it identifies precisely where evidence is thin or missing. Instead of allowing the AI to generate generic filler prose, this skill produces a structured TODO report that routes the user to specific upstream fixes—such as re-extracting paper details or expanding literature mapping—to ensure the final output is grounded and authoritative.
주요 기능
01Identifies evidence gaps and 'blocking missing' signals in research artifacts
02Provides a semantic routing matrix to map symptoms to root cause upstream fixes
03Generates a structured EVIDENCE_SELFLOOP_TODO.md with actionable fix paths
04Enforces a strict evidence-first contract to prevent the invention of facts
05Evaluates citation-backed anchors, evaluation metrics, and comparison density
0696 GitHub stars
사용 사례
01Automating the audit of research pipelines to ensure high-quality intermediate artifacts
02Unblocking academic survey writing when evidence bindings are insufficient
03Routing agents to rerun specific extraction skills when empirical data is missing