What problem does it solve? Systematic literature reviews require expanding a small set of known seed papers into a comprehensive candidate pool, but manually chasing citations forward and backward through the academic graph is slow and error-prone. This Skill automates citation snowballing, relevance search, and metadata lookup against the Semantic Scholar API (~200M+ papers) through a single stdlib-only Python CLI. ## Core Features & Use Cases - Citation Snowballing: Multi-hop forward (who cites a paper) and backward (what it cites) expansion from seed papers, with deduplication, provenance tracking (which seed, direction, hop, influential flag, citation intents), and filters for year, citation count, and fields of study. - Search & Lookup: Relevance search, boolean bulk search (up to ~1000 results), single-paper metadata by DOI/arXiv/S2 id, batch hydration of up to 500 ids, author search, and paper recommendations. - Disk Caching & Rate-Limit Handling: Every GET and batch POST is cached for 7 days by default, with ~1 request/second throttling and exponential backoff on 429/503 responses. - Use Case: Given five seed papers for a systematic review, run a 1-hop snowball in both directions restricted to influential citations, then hand the ranked candidate list (with DOIs) to the user for import into Zotero. ## Quick Start Ask Claude to snowball citations from your seed papers, for example: "Run a forward and backward snowball from DOI:10.1016/j.infsof.2008.09.009 with influential-only filtering and show me the top candidates."
