paper-download

name

paper-download

description

Use this skill when the user wants to download paper PDFs by DOI(s). Triggers: "下载这个/这几个 DOI的 PDF/全文", "把这个 DOI 列表全部下下来", "批量下 dois.txt", "download paper(s) for DOI(s) …", "bulk-download from this Web of Science / WoS savedrecs export". Routes each DOI through publisher TDM APIs (Wiley/Elsevier/Springer) then OA fallbacks (OpenAlex/Crossref/Unpaywall), and — when --use-browser-fallback is set — a final Playwright pass that reuses institutional cookies for ACS/RSC/IEEE/AIP/IOP/APS/Science. DO NOT use for paper *search* (use Semantic Scholar / arXiv instead), Zotero import / metadata sync, or downloading *all references of* a single paper (use the ref-downloader skill).

Paper Download — Agent Runbook

Slim runbook loaded when the skill is invoked. Configuration depth and human-facing reference live in the project docs:

• ../../../README.md — install + invoke (humans and agents)
• ../../../docs/BROWSER_FALLBACK.md — every browser-fallback knob
• ../../../docs/SUPPORTED_PUBLISHERS.md — per-publisher routing + tier table

Prerequisites (verify once per workspace)

1. auto_paper_download is importable. If not: pip install -e . from the repo root (add [browser] extra + playwright install chromium for browser fallback).
2. A .env exists in the cwd the scripts will run from. Minimum: CROSSREF_MAILTO (any real email). Missing creds don't error — they silently disable that path, so surface the warning to the user verbatim if creds are empty.

Pre-flight (per invocation)