literature-search-biorxiv

This skill lets an agent browse, filter, and fetch metadata for life-sciences, biology, and medical preprints from the bioRxiv and medRxiv archives. It solves a specific mismatch: bioRxiv/medRxiv expose a date-based archive API with no server-side keyword or author search, so naive attempts to search broadly quickly cause thousands of API calls, timeouts, and abuse blocks. The skill wraps the archive with rate-limited helper scripts and strong guidance to keep queries narrow and correct.

It provides two utility scripts run via uv: search_by_doi.py to fetch detailed JSON metadata for a single paper by DOI (the most reliable entry point), and search_by_dates.py to browse an explicit date range with optional category, keyword (AND/OR), and author filters applied locally after download. The documentation stresses core rules: always use the wrapper scripts (never custom curl), keep date ranges to 1-4 weeks paired with a category, redirect JSON output to a file, pass an include-abstracts flag only when needed to save context, and cite the source URLs of any papers used. It handles pagination automatically, enforces cross-process rate limits with backoff, validates subject categories, and prompts the user to review bioRxiv/medRxiv terms and per-paper licenses via a one-time notification file. Full-text PDF download is explicitly out of scope and delegated to a sibling EuropePMC skill.

Target users are researchers, scientists, and research-support agents who need to retrieve preprint metadata by DOI or survey recent preprints in a specific category and window. Typical use cases include resolving a citation's metadata and scanning a few weeks of a discipline for relevant new work.