Citation Reference Integrity Manager is a skill for developers and researchers building document workflows that require accurate, verifiable academic citations. It runs a five-pass quality control process on AI-generated text to detect and report reference integrity problems that standard LLM prompting routinely misses.
The skill performs integrity auditing to identify fabricated authors, non-existent journals, and malformed DOIs or PMIDs. It enforces citation style rules natively for AMA (11th edition), Vancouver (ICMJE), and APA (7th edition). A sync logic pass catches orphan citations — in-text markers with no corresponding bibliography entry — and uncited references that appear in the bibliography but nowhere in the body. Structural validation checks sequential numbering and correct journal abbreviations.
After the five passes complete, the skill produces a Reference Integrity Note summarizing the findings and a verdict. This gives pipelines a programmatic signal they can use to hold documents before they reach clients or journals. The skill is aimed specifically at workflows where citation hallucinations carry real consequences: medical, legal, or academic publishing, systematic review pipelines, and research report generation. It does not perform live network lookups to resolve DOIs or PMIDs against external databases; it validates structural correctness and cross-document consistency rather than confirming real-world existence of a source.