Paper Reference Authenticity Verification

Problem

References in academic papers can be more than formatting issues: they may include fabricated entries, mismatched authors, incorrect venues, stale DOIs, or predatory publications. This skill turns the question of whether a citation actually exists into a repeatable workflow. It extracts reference entries from endnotes, footnotes, or citation lists, then checks title, authors, venue, year, DOI, and publisher legitimacy.

How It Works

The workflow has five stages. First, it identifies the reference list and structures each item into fields such as number, title, authors, journal/conference/book, year, volume/issue/pages, and DOI. Entries with irregular formatting or only a URL are retained with explicit markers. Next, it searches across academic platforms by field and priority, including Google Scholar, Semantic Scholar, CrossRef, PubMed, IEEE Xplore, ACM Digital Library, arXiv, CNKI, Wanfang, Web of Science, Springer, and ScienceDirect. After a match is found, the skill checks five dimensions: existence, title accuracy, author accuracy, publication details, and venue legitimacy. It reports issues such as E-FAKE, E-DOI, E-RETRACT, E-PREDATORY, or W-UNVERIFIED with evidence and conclusions. For larger lists, it can prioritize irregular entries, obscure venues, and high-risk cases, while noting the sampling scope in the report.

Boundaries

The skill depends on public academic search results, DOI resolution, and journal or conference listings, so it is best suited for searchable publications such as journal articles, conference papers, theses, books, patents, and web resources. Paywalled content, private reports, very recent preprints, and region-restricted sources should be marked as unverifiable rather than immediately labeled fake. For Chinese literature, prioritize CNKI and Wanfang; predatory journal checks can be cross-checked against Beall's List and COPE resources.