PlugWiki

nature-downloader (nature-skills): a Claude skill that downloads paper full text through open access, publisher APIs and your library login

nature-downloader makes Claude fetch the full text of papers you name, using open-access sources, publisher APIs, CNKI or your own library login.

Installs
12.4K
on skills.sh
Stars
51.7K
on GitHub
Files
60+
besides SKILL.md
Updated
Nov 2026
last repo commit

Give Claude a list of DOIs or exact titles and it runs a batch script that tries the lawful routes in order. Chinese papers go to CNKI. For Elsevier, Springer Nature and IEEE it tries the publisher API first, then open-access sources such as PMC, Unpaywall and arXiv. Other publishers go to open access first, then to your institution's access in your own Chrome.

Before it downloads anything, Claude asks whether you also want Supporting Information. It needs Node.js 22+ and Python 3. The CNKI, institutional and Supporting Information routes also need Chrome with remote debugging turned on and you logged in. Each run writes a manifest.json that lists what came from where.

Good for

  • Researchers who need the PDFs for a list of DOIs or titles
  • Chinese-language researchers who pull papers from CNKI through their university
  • Students whose library gives access through CARSI, EZproxy or WebVPN

Install nature-downloader

Claude Code

  1. 1

    Download the skill folder from https://github.com/yuan1z0825/nature-skills/tree/HEAD/skills/nature-downloader.

  2. 2

    Put it at ~/.claude/skills/nature-downloader/ so that SKILL.md sits directly inside.

  3. 3

    Start a new session; the skill loads when its description matches the task.

The Claude app ZIP upload is not offered here, because the skill needs Node.js, Python and a local Chrome. The skill also expects a separate web-access skill for its Chrome proxy.

Built from the package data we hold. Not yet tested by us.

What's in the skill

When it loads

Use when a user needs lawful academic full text, CNKI institutional access, English OA retrieval, publisher API access, institutional browser fallback, or supporting information downloads.

Sections in SKILL.md

  1. Download Intake and First-Run Configuration
  2. Paid Library Resource Configuration
  3. API-First and Open-Access Fallback
  4. Publisher API Credentials
  5. Resource URL Triage
  6. Institution-Specific Domains
  7. Boundaries
  8. Preconditions
  9. Batch Scope
  10. Status Categories
  11. Start Browser Control
  12. Fast Batch Path (default for 2+ papers — fast & token-efficient)
  13. Advanced Browser and Delivery Routes

Key facts

Author
yuan1z0825 (GitHub)
Repo
yuan1z0825/nature-skills (skills/nature-downloader)
License
Apache-2.0
Installs on skills.sh
12,385 as of 2026-11-02
GitHub stars
51,696 (whole repo) as of 2026-11-02
Last repo commit
2026-11-01
Folder contents
71 files including SKILL.md: 39 Node scripts, 12 Python files, 7 Markdown files, config data, example templates and tests
Needs
Node.js 22+ and Python 3; Chrome with remote debugging for CNKI, institutional and SI routes
Python packages
pdfplumber, pypdf, PyYAML, jsonschema (requirements.txt)
skills.sh audits
Gen Agent Trust Hub pass, Socket fail, Snyk pass

What it includes

  • Asks first whether you want Supporting Information, then runs with --si or --no-si.

    SI confirmation gate

  • Picks CNKI, publisher API, open access or institutional browser by language and publisher.

    Route selection

  • Downloads by DOI list, title, PDF URL or Web of Science topic, with a manifest.

    Runs code · batch_download.mjs

  • Searches PMC, Unpaywall, publisher OA pages, arXiv and other repositories.

    Runs code · Open-access lookup

  • Uses an Elsevier, Springer Nature or IEEE key; falls back to open access on failure.

    Runs code · Publisher API download

  • Downloads Chinese papers through your logged-in Chrome session.

    Runs code · CNKI route

  • Uses your library login in Chrome, only after you confirm when an API failed.

    Runs code · Web access fallback

  • Infers your library route from a resource URL and saves it as a config file.

    Runs code · configure_school.py

  • Saves, validates and deletes publisher keys and a contact email.

    Runs code · configure_credentials.py

  • Records one of 29 statuses per paper in manifest.json.

    Write · Status codes and manifest

  • Finds and downloads supplement files from the article page when you ask.

    Runs code · Supporting Information

  • Checks each file is a real PDF and names it for delivery.

    Files · Verification and naming

From the official docs

Permissions

High risk

This skill runs scripts, controls your logged-in Chrome and saves publisher API keys on your disk, so a mistake touches your institutional account. Confirm the paper list yourself, and stop if your library or publisher shows a security warning.

What it can do

  • Read your data
  • Create and change things
  • Delete things
  • Read-only mode

Watch out for

  • It drives your logged-in Chrome session

    For CNKI, institutional access and Supporting Information it attaches to your Chrome through remote debugging. SKILL.md says it must not read cookies or session files. Download only papers you are entitled to.

  • It stores publisher API keys on disk

    `configure_credentials.py` saves keys to ~/.config/lit-dl/credentials.json. SKILL.md says a key you paste into chat is saved as sent, so use the hidden prompt instead.

  • It tries to pass robot checks

    SKILL.md tells Claude to try visible slider, checkbox and Verify controls up to twice per tab. Check your library and publisher terms first, since automated access can breach them.

  • It sends your contact email to Unpaywall and Crossref

    The open-access lookup calls api.unpaywall.org with your contact email, and the Crossref metadata lookup puts it in a mailto parameter or User-Agent header. You set it with `configure_credentials.py contact-email` or UNPAYWALL_EMAIL.

  • It installs and runs local code

    It runs Node.js 22+ and Python scripts, and its requirements.txt lists pdfplumber, pypdf, PyYAML and jsonschema. It also expects a separate web-access skill for the Chrome proxy.

Technical details
Talks toapi.elsevier.com. api.springernature.com. ieeexploreapi.ieee.org. api.crossref.org. api.unpaywall.org. www.ncbi.nlm.nih.gov. ftp.ncbi.nlm.nih.gov. www.ebi.ac.uk. export.arxiv.org. arxiv.org. doi.org. kns.cnki.net. www.webofscience.com. 127.0.0.1:3456 (local Chrome proxy). your library and publisher sites
Sign-in and keysPublisher keys are saved in ~/.config/lit-dl/credentials.json with mode 0600, according to SKILL.md. Library access reuses your logged-in Chrome session.

Questions

What does nature-downloader do?

It downloads the full text of papers you name by DOI, PMID or exact title. It tries open-access sources and publisher APIs first, and uses CNKI or your institution's login in Chrome when it has to. Each run saves a manifest with the route, file hash and result for every paper.

How do I install it?

Use the install box above. It is one folder in the nature-skills repo. The skill also needs Node.js 22+ and Python 3. If you use the Python helpers, install the four packages in its requirements.txt.

Does it need an API key?

Not for open access. Elsevier, Springer Nature and IEEE downloads use a key from that publisher, which you save once with the skill's configure_credentials.py script. SKILL.md says not to paste keys into chat, though it saves a key you send voluntarily.

Will it bypass a paywall?

No. SKILL.md says it uses only lawful open access, publisher APIs and your own institutional access, and that it must not bypass paywalls, DRM or two-factor authentication. If your library has no entitlement, the paper is marked library_no_permission.

Does it solve CAPTCHAs for me?

Partly. SKILL.md tells Claude to try a visible slider, checkbox or Verify button in your Chrome session, at most twice per tab, before asking you. It hands over to you for image choices, QR codes, SMS codes and two-factor prompts.

Is it safe?

It drives your logged-in Chrome session and saves publisher API keys on your disk, so review what you approve. SKILL.md says it must not export cookies or ask for institutional passwords, and it limits downloads to the list you confirm. Use it only where your institution's terms allow automated downloading.

What is the web-access skill it mentions?

For the browser route, SKILL.md expects a separate web-access skill that provides a local Chrome proxy on port 3456. That skill is not in this repo, and nature-downloader does not install it for you.

Does it work in Codex?

Yes. The nature-skills README documents installing the skill for Codex, and SKILL.md lists .codex\skills and .agents\skills as install locations. The Claude Code path in the box above puts it in ~/.claude/skills.

Information

updated Nov 2, 2026

Repository
yuan1z0825/nature-skills
Folder
skills/nature-downloader

yuan1z0825/nature-skillsskills.sh

Sources

Docs, issue trackers and discussions this page is built from.

Show all 5
  1. nature-skills README · github.com
  2. nature-downloader SKILL.md · github.com
  3. delivery-verification-and-failures.md · github.com
  4. requirements.txt · github.com
  5. nature-downloader on skills.sh · skills.sh

Updated Nov 2, 2026.