What problem does it solve? Academic papers often contain fabricated DOIs, tampered metadata, withdrawn papers, mismatched authors, and missing key references that manual review misses. This Skill runs a three-phase verification pipeline over a paper directory to detect fake references, confirm each citation actually supports the claim it is attached to, and find important literature that was never cited. ## Core Features & Use Cases - Phase 1 — Existence check: Verifies every DOI against doi.org, Crossref, and Semantic Scholar, distinguishing fully fabricated references from tampered DOIs, replacing fake entries, and running a D10a regression check so no orphan \cite keys remain in the .tex file. - Phase 2 — Appropriateness check: Reads each reference PDF in full (via pymupdf/pdfplumber), compares title, author, year, and content against the citation context, and detects withdrawn papers, cross-domain DOIs, fabricated author names, and Nature/Springer download mix-ups using pdfinfo. - Phase 3 — Completeness check: Independently searches Semantic Scholar for high-citation literature on the paper's topics and classifies omissions as critical, suggested, or optional based on citation count and topic-match thresholds. - Use Case: Before submitting a CRISP-DM methodology paper, point the Skill at the paper directory; it flags a fabricated SMOTE reference whose DOI resolves to a finance paper, repairs a tampered Kapoor 2023 DOI, and recommends adding the Martinez-Plumed 2021 CRISP-DM retrospective. ## Quick Start Run citation verification on my paper directory papers/my-paper and produce the phase1, phase2, phase3, and overall verification report with a repaired bib file.
