What problem does it solve? Scientific papers often contain unverifiable numbers, fabricated citations, fake DOIs, and inconsistent values across sections. This Skill enforces a fixed multi-layer quality gate pipeline (L0 direction check, L0.5 data honesty, G1-G7 pipeline gates, L4 content review) so every claim in a paper is traced to a source file before submission. ## Core Features & Use Cases - Layered Gate Architecture: Runs G1-G7 gates (identity, compilation, citation integrity, constitution compliance, citation quality, impact mapping, content review) plus an L0.5 data honesty gate with one-vote veto power. - Deterministic Runner: All checks execute through quality-gate-runner.py in full or fast mode, producing a structured quality_report.json with per-gate scores, issues, and P0/P1/P2 severity classification. - Repair Loop: Failed gates trigger prioritized fixes (patch paper.tex, clean bib files, recompile LaTeX) with up to three retry cycles before escalating to human review. - Use Case: Given a LaTeX paper directory with experiment JSON outputs and a references.bib file, run the full gate to verify every numeric claim traces to a source file, every DOI resolves, and compilation succeeds with zero undefined citations. ## Quick Start Run the quality gate on my paper directory outputs/papers/pima-crispdm in full mode and generate the quality report.
