What problem does it solve?
Literature reviews often contain mismatched or hallucinated citations where the cited paper never actually did what the text claims. This Skill checks every citation in a {topic}_review.tex file against the real BibTeX metadata and PDF content, then minimally rewrites only the sentences with fatal citation errors.
Core Features & Use Cases
- Semantic Citation Verification: The host AI checks each citation against PDF excerpts and BibTeX abstracts, classifying issues as fake, wrong, contradictory, weak, or overstated citations.
- Minimal Rewriting with Priority Levels: Only P0 fatal errors (fake/wrong/contradictory citations) are rewritten; P1 issues are warnings only and P2 style issues are never touched, preserving all LaTeX commands.
- Deterministic Helper Scripts: Python scripts parse LaTeX paragraphs, extract citation contexts, build a structured
ai_alignment_input.json, and render PDF/Word output by reusing the research-literature-review skill.
- Use Case: After generating a systematic review, run this Skill to catch hallucinated references, produce an
ai_alignment_report.md audit trail, and regenerate the corrected PDF and Word files.
Quick Start
Ask the AI to use research-citation-check to verify and fix the citations in the review tex file inside your project directory.