What problem does it solve? LLMs and manual errors produce citations that look plausible but do not exist, or attach real DOIs to fabricated papers. This Skill checks every entry in a bibliography against authoritative databases so you can catch phantom, chimera, and corrupted references before journal submission. ## Core Features & Use Cases - DOI existence and metadata verification: Resolves each DOI against CrossRef (the DOI registrar) and compares stated title, year, and first author against the registered record, with OpenAlex and PubMed as fallback sources for title search. - Broad input support: Accepts BibTeX (.bib), LaTeX (.tex/.bbl), Word (.docx), PDF, plain-text/Markdown reference lists, and Google Docs URLs, with an arXiv API fallback for DataCite-registered arXiv DOIs. - Structured TSV report: Outputs per-citation status (VERIFIED, PHANTOM, CHIMERA, CORRUPTED, SUSPICIOUS, NOT_FOUND, UNVERIFIABLE), similarity scores, flags, and a sidecar metadata JSON with input hash and per-status counts. - Use Case: Before submitting a manuscript, export your Zotero library as refs.bib and run the validator to find any hallucinated or mis-attributed citations an AI assistant may have inserted. ## Quick Start Ask the assistant to validate the citations in your exported refs.bib file against CrossRef and report any hallucinated references.
