Paper Search and Literature Lookup

Problem

When doing research or technical investigation, you often need to confirm whether a direction has enough literature, then check titles, authors, years, citation counts, DOIs, and full-text links one by one. If this is done by switching between browsers and multiple academic sites, it can miss sources and is inconvenient for downstream scripting. This skill compresses common literature-lookup needs into a single natural-language query, useful for early screening before topic selection, review writing, related-work checks, or engineering research.

How It Works

• Intent input: accepts keywords, research fields, authors, and filters, such as “LLM lightweighting,” “Zhang San deep learning,” “computer vision multimodal,” or “highly cited papers in the last five years.”
• Multi-source retrieval: queries open academic APIs such as OpenAlex, arXiv, and Crossref in real time, without relying on a local static JSON library, and covers Chinese and English journals, SCI/EI, top conferences, and preprints.
• Structured output: returns the paper title, publication year, author team, journal or conference source, citation count, DOI, and accessible full-text links such as PDFs or official pages.
• Programmatic use: run node search.mjs "LLM lightweighting" to search, use --format=json for structured output, and use --limit=3 to control the number of results.

Boundaries

It is best for paper discovery, metadata lookup, and citation positioning, helping you compare candidate papers and obtain DOIs or full-text entries. It does not manage a bulk reference library, parse PDFs deeply, extract figures, generate semantic literature reviews, or convert citations to BibTeX/EndNote formats; those require other tools.