Orchestrates parallel searches across Semantic Scholar, arXiv, OpenAlex, and CrossRef, then merges and deduplicates results into structured JSONL with BibTeX entries. The real workflow here is query expansion into 2-4 variants, followed by citation-weighted ranking that balances recency, venue quality, and relevance. Semantic Scholar is treated as primary for ML/AI papers since it includes proper BibTeX, while arXiv catches the last three months of preprints. Nice touch: it can download arXiv source LaTeX files directly. The venue tier system is opinionated but sensible, clearly marking NeurIPS/ICML as top tier and flagging preprints. Use this when you need a proper literature survey with mergeable citations, not just Google Scholar copypasta.
npx skills add https://github.com/lingzhi227/agent-research-skills --skill literature-search