Nature Citation Skill: Find Nature, Science and Cell Papers for Your Claims
Makes Claude split manuscript text into claims, find supporting papers in Nature, Science and Cell journals, and export them for EndNote or Zotero.
- Installs
- 16.1K
- on skills.sh
- Stars
- 51.7K
- on GitHub
- Files
- 16
- besides SKILL.md
- Updated
- Nov 2026
- last repo commit
Nature Citation makes Claude break your text into short claims, search Crossref and PubMed for papers in Nature Portfolio, Science and Cell Press journals, and grade how well each paper supports each claim. It then writes one reference-manager file, RIS by default, or ENW or Zotero RDF if you ask.
The README lists it as Beta. It runs a Python script that sends search queries to the internet, so do not paste text you cannot share with those services. Treat every candidate as an option and read the abstract before you cite it.
Good for
- Researchers adding citations to an introduction or discussion
- Authors who need supporting papers only from Nature, Science, Cell and their sister journals
- People who already have DOIs and want a clean reference-manager file
Install nature-citation
Claude Code
- 1
Copy this prompt.
Install the nature-citation skill in Claude Code. Run this command: npx skills add Yuan1z0825/nature-skills --agent claude-code --skill nature-citation --yes --copy Then check that it loaded and tell me how to use it.
- 2
Paste it into Claude Code. It shows the change and asks before running it.
Requires Node.js 18+.
nature-citation does not use the nature-shared package (its manifest says so), so it installs alone. The script needs Python and the defusedxml package; the README says the installer does not install Python dependencies for you. The README's own Claude Code route is a git clone plus a subagent or slash command wrapper; the command here is the README's Codex command with the agent changed, so it is generated, not quoted.
Built from the package data we hold. Not yet tested by us. Source.
What's in the skill
When it loads
Find and verify Nature/CNS-family literature supporting manuscript claims, with claim-to-source mapping and reference-manager export. Use for Nature系列引用、CNS支撑文献、分段补引用 when this journal scope is requested; use broader literature search for unrestricted sources.
Sections in SKILL.md
- Routing protocol
- 1. Load the manifest and the core layer
- 2. No content axis — confirm scope and language inline
- 3. Run the workflow
- 4. Reach for references only when needed
Key facts
- License
- Apache-2.0
- Installs on skills.sh
- 16,068 (Nov 2, 2026)
- Last commit
- Nov 1, 2026
What it includes
Splits into citable claims with stable IDs like S001
Uses paragraph breaks, then sentences for long paragraphs. Over about 10 segments it switches to batches.
Try: “Add Nature-series citations to this introduction paragraph.”
Read ·
1. Segment the textExtracts the core claim, its type and 2 to 4 English search queries
Claim types include mechanism, association, method, clinical and background.
Read ·
2. Parse each claimRuns scripts/nature_citation.py against Crossref and PubMed
Scope flags are nature, cns, cns plus sister journals and flagship only. Year filters and batch size are also options.
Runs code ·
3. Search candidatesStrong, partial, background, contradictory or metadata-only
A metadata-only candidate is never cited as support until the abstract or publisher page is checked.
Read ·
4. Grade supportRIS by default, or ENW or Zotero RDF
An author-integrity check blocks records with missing or surname-only authors, then refetches by DOI or PMID.
Write ·
5. Export one fileOptional HTML browser, JSON, TSV or Markdown with --with-artifacts
Useful for long or ambiguous texts.
Write ·
6. Review artifactsClaim-to-source table with insertion points, DOIs and support notes
Shows the scope and date limits used, plus open gaps.
Read ·
7. Final report
From the official docs
Permissions
Medium risk
The script sends search queries built from your text to Crossref and PubMed, and writes export files to a folder on your machine. Do not use it on unpublished text you cannot share.
What it can do
- Read your data
- Create and change things
- Delete things
- Read-only mode
Watch out for
Your claims leave your machine as search queries
The script builds queries from your text and sends them to Crossref, and to PubMed when you give PMIDs. Use short claims, not confidential data, if the manuscript is unpublished. github.com
The script needs a third-party Python package
The script imports defusedxml (line 24), which is not in the standard library. The README says the installer does not install Python dependencies for you, and does not list defusedxml for this skill. Install it yourself, for example in a virtual environment. github.com
Optional email and Firecrawl key
You can pass --mailto and --ncbi-email, which are sent to Crossref and NCBI as contact details. The core guide also mentions Firecrawl with a FIRECRAWL_API_KEY for pages that need JavaScript. github.com
Candidates are not proof
The skill says never to present a paper as support only because its title is related, and to check the abstract before citing a metadata-only candidate. Still read the paper. github.com
An override can export incomplete author lists
workflow.md blocks export when an author name is incomplete and allows --allow-incomplete-authors only as an explicit emergency override. Do not use it for a submission. github.com
Technical details
| Talks to | api.crossref.org (Crossref search and DOI records). eutils.ncbi.nlm.nih.gov (PubMed author records for PMIDs) |
|---|---|
| Scopes |
|
Questions
What does the nature-citation skill do?
It finds supporting references for your manuscript claims, limited to Nature Portfolio, Science and Cell Press journals, explains how strongly each paper supports each claim, and exports a RIS, ENW or Zotero file.
How do I install it?
Use the install box at the top of this page. Its manifest says it does not use the nature-shared package, so it can be installed on its own. It is a Python script, so also have Python and the defusedxml package ready.
Does it need an API key?
No key is required for the script, which calls the public Crossref and PubMed endpoints. A Firecrawl key is mentioned only as an option for pages that need JavaScript.
Is my manuscript text sent anywhere?
Search queries built from your claims go to Crossref, and to PubMed when you give PMIDs. Your text also goes to the AI model you use. Avoid pasting confidential results.
Which journals does it search?
You choose a scope: Nature series, CNS (Cell, Nature, Science plus sister journals), or flagship only. It checks the journal against a publisher-family list and does not accept titles that only contain the word "nature".
Is it safe to rely on the citations?
Treat them as candidates. The README says candidate papers are support options, not a guarantee that the final citation is appropriate. Read each paper before you cite it.
How is it different from nature-academic-search?
The README describes nature-academic-search as the broader literature search and citation-metric audit tool. Nature Citation is strict about journal scope and focuses on claim-to-source mapping and export.
Information
updated Nov 2, 2026
- Repository
- yuan1z0825/nature-skills
- Folder
skills/nature-citation
Sources
Docs, issue trackers and discussions this page is built from.
Show all 9Hide
- Nature Skills README, installation · github.com
- skills CLI commands · github.com
- manifest.yaml · github.com
- github.com · github.com
- github.com · github.com
- Nature Skills README (index) · github.com
- nature_citation.py · github.com
- workflow.md · github.com
- script-usage.md · github.com