citation-verification

Search for each paper with WebSearch, confirm it exists on Google Scholar, check that title, authors, year, and venue match, then copy the BibTeX directly from Google Scholar. Verify at the moment you add the citation rather than after writing.

Never trust citations generated from memory, since AI-generated references have roughly 40% error rates. Verify every citation through WebSearch and Google Scholar before adding it to your bibliography, and mark unverifiable ones as [CITATION NEEDED].

Yes, the skill includes reference Python scripts using CrossRef, arXiv, and Semantic Scholar APIs for batch or automated verification. However, the primary recommended workflow uses WebSearch and Google Scholar rather than these API scripts.

Do not guess or fabricate the reference. Check spelling, try alternate queries, look on arXiv or via DOI, and if it still cannot be found, mark it as [CITATION NEEDED] and clearly notify the user with the reason verification failed.

Yes, when citing a specific claim, the workflow requires accessing the paper, searching for relevant keywords, and confirming the claim actually appears, recording the section or page where it is found.