nature-downloader

Run scripts/batch_download.mjs with --dois followed by a comma-separated DOI list, an --out directory, and exactly one of --si or --no-si. The script routes each paper through publisher APIs, open access, CNKI, or institutional access and writes a manifest.json with per-paper results.

Pass a Chinese title with --title and the downloader automatically uses the CNKI route through your authenticated browser session. You can force PDF-only output with --cnki-format pdf or supply a library-provided CNKI entry with --cnki-url.

Yes, configure API keys with scripts/configure_credentials.py for Elsevier, Springer Nature, or IEEE, and the downloader tries the publisher API first for those papers. If the API lacks entitlement, it falls back to open-access sources before asking about institutional Web Access.

No, it never bypasses paywalls, DRM, or two-factor authentication and never reads cookies or passwords. It makes at most two bounded attempts on simple visible verification widgets, then hands the tab to the user for manual completion.

Every download requires an explicit supporting information choice, so the script exits with si_confirmation_required when neither --si nor --no-si is provided. Re-run with one of those flags; the choice applies to the entire batch.

The paper is marked library_no_permission in the manifest when the resolver or publisher page explicitly denies institutional entitlement. The skill reports this plainly instead of retrying as if it were a temporary network problem.