SuperAPI AI Directory
Partner with us中文AI newsFavorites
Skillsnature-downloader

nature-downloader

Writing & research

View repository

Updated 10-16 01:30

About this skill

Nature Literature Downloader

Retrieve a definite paper list through lawful routes. Use nature-academic-search first when the target papers are not yet known.

Intake

Establish the identifiers, output location, desired format, and whether to fetch Supporting Information (SI). An explicit SI or main-text-only request is enough; otherwise ask once for the whole batch before downloading files.

Read manifest.yaml, then load only the references and scripts matching the selected route.

Route

  1. Normalize DOI/title/URL and identify language and publisher.
  2. Chinese literature: use the configured CNKI route.
  3. Supported publisher with configured API entitlement: try the API, then lawful OA when the API does not yield usable full text.
  4. Other English literature: try lawful OA first.
  5. Use institutional browser access only with the user's existing authorization. Confirm before switching to it after an API-plus-OA failure.
  6. For batches, prefer scripts/batch_download.mjs; use the focused helper named by the manifest for configuration, browser fetching, or verification.

Non-negotiable boundaries

  • Do not bypass paywalls, DRM, access controls, CAPTCHAs, Turnstile, reCAPTCHA, OTP, QR approval, passkeys, or identity checks. Keep the tab open and hand those steps to the user.
  • Do not read, export, log, or request passwords, cookies, browser storage, session tokens, OTPs, or recovery codes. Save publisher API keys only through the bundled secret-input flow; never echo them.
  • Use only a browser session the user has authorized. Do not claim that a fresh profile proves missing institutional entitlement.
  • Do not invent PDF, SI, resolver, or publisher URLs. Do not turn a failed lawful route into an unauthorized-mirror search.
  • PDF, HTML, XML, CAJ, archives, and attachments are distinct deliverables. Report the format actually obtained.

Load references/access-safety-and-handoff.md before any institutional or browser-authenticated route.

Validate and record

For every target:

  • verify content type and signature (%PDF for PDF), non-empty page/content, and title or DOI match;
  • retain the source URL, access mode, timestamp, format, file hash, SI status, and canonical status from scripts/lib/status-codes.mjs;
  • distinguish success, unavailable-but-readable HTML, no entitlement, authentication handoff, ambiguous metadata, and typed failure.

Load references/si-format-and-file-validation.md when SI or non-PDF formats are involved.

Completion

Finish when every requested paper has either a validated artifact or an explicit typed failure, and the batch manifest is complete. Report user action only for the papers that still require it. Never report a download as successful solely because a script exited without error.