What problem does it solve?
AI-generated academic papers often contain fabricated empirical results, unsupported numeric predictions, or typography that silently breaks XeLaTeX rendering. This Skill provides a deterministic, fail-closed audit that blocks manuscript compilation when such publication-safety violations are detected.
Core Features & Use Cases
- Evidence-Free Claim Detection: Rejects completed-finding claims, reported significance statistics, and concrete predicted outcome numbers (in English and Chinese) when no user-supplied evidence exists, while allowing explicit hypotheses, planned thresholds, and declared metric targets.
- Caption-Level Auditing: Independently checks every figure and table caption for categorical outcome claims such as "the proposed method achieves lower cost" when evidence is absent.
- LaTeX Typography Enforcement: Blocks literal Unicode Greek glyphs and Latin-style en/em dashes, requiring named LaTeX math macros and
-- / --- punctuation to avoid missing-glyph degradation in XeLaTeX.
- Use Case: As an internal gate in an automated paper-writing pipeline, it runs after LaTeX sanitization and immediately before compilation, exiting non-zero with a list of blockers whenever upstream length or citation gates fail or the manuscript makes unsupported claims.
Quick Start
Run the paper quality gate on the generated manuscript package to verify all length, citation, and evidence checks pass before compiling the paper.
