Download PDFs for a Paper List

Given a list of academic papers (titles, DOIs, arXiv IDs, etc.) and a target directory, download the source PDFs.

The skill works in two phases:

1. Fast path — fetch open-access PDFs directly (arXiv, PMC, anything surfaced by Semantic Scholar's openAccessPdf.url and served as a real PDF). Pure stdlib, no browser, ~1s per paper.
2. Browser path — for everything else (paywalled publishers, Cloudflare- protected hosts, JS-redirected logins), launch Chromium via Playwright with a persistent user profile so the user can complete institutional SSO / publisher login in a real browser window. Sessions persist across runs — the user logs in to each publisher once.

Real publishers (Elsevier, Springer/Nature, Wiley, ACS, IEEE…) fingerprint TLS and JS environments, so a urllib/curl-with-pasted-cookies approach fails on most of them. A real browser is what works.