What problem does it solve? Growing a literature review beyond its seed papers requires manually chasing citations and judging each new paper's relevance, which is slow and inconsistent. This Skill automates recursive citation-graph expansion and relevance screening so a corpus grows outward from real citation edges instead of guesswork. ## Core Features & Use Cases - Recursive citation crawling: Follows backward references and forward citations via OpenAlex, falling back to Semantic Scholar for unindexed works like new arXiv preprints, bounded by configurable hop, fanout, node-budget, and date-floor limits. - Batch relevance classification: Sends unclassified papers to free OpenRouter models in batches, producing a 5-point relevance verdict plus a strategy-family tag that is fuzzy-deduplicated against the existing vocabulary. - Graph export and review hand-off: Writes graph_data.json for citegraph.html, a ranked candidates_for_review.md for manual promotion into the bibliography, and a families.md tag vocabulary report. - Use Case: After an initial deep-research pass yields 7 seed papers, run the crawl to discover 300 related works, classify each for relevance to your research question, and review the top candidates for inclusion in your bibliography. ## Quick Start Ask the AI to expand the citation graph of the existing academic-research corpus and classify the newly discovered papers by relevance.
