literature-review

Define a research question using the PICO framework, then search at least three databases such as PubMed, bioRxiv, and Semantic Scholar. Aggregate results with search_databases.py to deduplicate by DOI, screen titles and abstracts against documented criteria, and synthesize findings thematically.

Run verify_citations.py on your markdown file to extract every DOI, confirm each resolves through the DOI handle API, and pull authoritative metadata from CrossRef. The script outputs a verification report plus correctly formatted citations in APA or Nature style.

The skill supports APA 7th edition, Nature, Vancouver, Chicago author-date, and IEEE styles. Detailed formatting rules are in references/citation_styles.md, and generate_pdf.py accepts a --citation-style flag to apply one consistently.

Run generate_pdf.py on your markdown file, which invokes pandoc with the xelatex engine to produce a formatted PDF with table of contents and numbered sections. Both pandoc and a TeX Live installation must be present on the system.

Use field-normalized metrics from OpenAlex such as FWCI and citation_normalized_percentile rather than raw citation counts, since citation rates differ by an order of magnitude between disciplines. A FWCI of 1.0 represents the field average, and 5 or above indicates highly influential work.

Single-database searches miss preprints, cross-disciplinary work, and non-biomedical literature, making the review incomplete and irreproducible. The skill requires a minimum of three complementary databases including a preprint server like bioRxiv or arXiv.