nature-paper-card Skill for Claude: What It Does and How to Install It

Makes Claude deep-read one scientific paper into a 16-section Paper Card that ties each claim to its evidence.

Installs: 10.9K; on skills.sh
Stars: 51.7K; on GitHub
Files: 21; besides SKILL.md
Updated: Nov 2026; last repo commit

This skill turns one paper into a Markdown Paper Card with 16 numbered sections, from basic information and the research question to the experiment-to-claim evidence chain, limits, critique and research ideas. It picks one paper type (methods, discovery, resource, clinical, materials or review) and marks anything it cannot support as not assessable.

For a PDF it runs the bundled prepare_paper.py, then checks the finished card with audit_paper_card.py. It needs Python 3 and nature-shared. It does not translate the full paper or write a formal peer review, and it matches your language by default.

Good for

Researchers who need to read a paper closely
Students preparing a group-meeting or reading-group talk
Anyone checking what a paper does and does not prove
Researchers looking for follow-up ideas in a paper's limits

Install nature-paper-card

Claude Code

Copy this prompt.

Install the nature-paper-card skill in Claude Code. Run this command:

npx skills add Yuan1z0825/nature-skills --skill nature-paper-card --skill nature-shared

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.

Pick one route. nature-paper-card loads a file from the nature-shared folder (its manifest lists ../nature-shared/core/terminology-ledger.md), and the README says to install nature-shared with the skills that use it. The skill runs two bundled Python scripts, so it needs Python 3, and PDF input needs PyMuPDF.

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

What's in the skill

When it loads

Build a structured deep-reading Paper Card for one scientific paper, analysing methods, experiment-to-claim evidence, limitations, and research ideas. Use for 论文精读卡、方法拆解、证据链分析; not full-paper bilingual translation or formal peer review.

Sections in SKILL.md

Routing protocol
1. Load the manifest and core layer
2. Establish the source boundary
3. Classify the paper type
4. Build the evidence base before drafting
5. Generate the fixed Sections 01-16 Paper Card
6. Run groundedness QA
Script red lines
Relationship to adjacent skills

Key facts

Author: Yuan1z0825 (repo owner)
License: Apache-2.0 (crawl data; the README badge and .codex-plugin/plugin.json also say Apache-2.0)
Installs on skills.sh: 10,875 (Nov 2, 2026)
GitHub stars: 51,696 for the whole Yuan1z0825/nature-skills repo (Nov 2, 2026)
Last repo commit: Nov 1, 2026
What it includes: 22 files including SKILL.md: 2 Python scripts, 2 test files, 6 paper-type fragments, 3 core files, 3 references, a manifest, an eval file, 2 README files and agents/openai.yaml
Needs: Python 3. prepare_paper.py needs PyMuPDF for PDF input, and a paper PDF, text, DOI or an existing nature-reader source map
Needs alongside: The nature-shared folder in the same repo. The manifest loads its terminology ledger every run
Works in: Claude Code, Codex, OpenClaw, OpenCode and Hermes, per the repo README install section

What it includes

Reads manifest.yaml and four always-loaded files
Loads the terminology ledger from nature-shared plus principles, workflow and output contract. SKILL.md says not to generate the card from the router alone.
Read · Load the manifest and core
Full paper, text only, abstract only, or a nature-reader map
For a PDF or source-map JSON, SKILL.md makes the bundled script mandatory, and partial input gives a visibly partial card.
Read · Set the source boundary
Turn a PDF or source map into source_bundle.json
Run by its resolved path. It can also render page images into a folder for visual review.
Try: “Use nature-paper-card to deep-read this PDF and make a complete Paper Card.”
Runs code · prepare_paper.py
page-grounded, structure-grounded or source-limited
Decides whether citations use PDF page numbers, figure and section pointers, or only abstract-level pointers.
Read · Pick a locator mode
methods, discovery, resource, clinical, materials or review
Loads one primary paper-type file, and at most one secondary for a hybrid paper.
Read · Classify the paper type
Evidence inventory and claim-evidence matrix
Lists metadata, claims, methods, every main figure and table, results and author-stated limits before drafting.
Read · Build the evidence base
Sections 01 to 16 in a fixed order
Writes paper-card.md by default. Section 16 holds research ideas, framed as hypotheses.
Write · Write the Paper Card
Check structure, locators and evidence grounding
Writes audit-report.json. SKILL.md treats audit errors as blockers.
Runs code · audit_paper_card.py
No replacement or patched scripts in a normal run
Claude must not write inline Python, a one-off extractor or invent page numbers when preparation fails.
Read · Script red lines

From the official docs

Permissions

Low risk

What it can do

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

Watch out for

Runs bundled Python scripts on your files
SKILL.md tells Claude to run both scripts by their resolved path and never to write inline Python or patch them in a normal run. Keep your paper files in a work folder you are happy to write to. github.com
PDF input needs PyMuPDF
If PyMuPDF is missing, prepare_paper.py prints that it is required and shows a pip install command. Install it yourself in a virtual environment. github.com
External search may be used for some sections
SKILL.md allows external search only for Section 04, Section 15, bibliographic verification or an explicit novelty check, and asks Claude to record the context mode. github.com

Security advisories

No published security advisories as of Nov 2, 2026.

Technical details

Turning on read-only	No read-only mode. Use version control.
Sign-in and keys	The bundled scripts read no environment variables and SKILL.md mentions no account or key.
Scopes	Claude can run python prepare_paper.py and python audit_paper_card.py from the skill folder Scripts read your paper (PDF or source-map JSON) and write source_bundle.json, audit-report.json and optional rendered pages Claude writes paper-card.md in the work folder

Questions

What does the nature-paper-card skill do?

It reads one paper and writes a 16-section Paper Card. The card covers the question, method, evidence for each claim, limits and testable ideas, with source pointers.

How do I install nature-paper-card?

Use the install box at the top of this page. It installs the skill together with nature-shared, because the skill's manifest loads a file from that folder. It has no Claude app upload because the skill runs Python.

Does it need an API key?

Neither bundled script reads environment variables. SKILL.md allows external search only for Section 04, Section 15, bibliographic verification or an explicit novelty check.

How is it different from nature-reader and nature-reviewer?

SKILL.md sends full-text bilingual reading and source maps to nature-reader, and formal reviewer-style assessment to nature-reviewer. This skill makes a single-paper deep-reading card and says not to silently switch to those outputs.

Does it work when I only have an abstract?

Yes, in a limited way. SKILL.md has a source-limited mode that avoids page numbers and marks unsupported sections "Not assessable from supplied material".

Which language does the card use?

SKILL.md says to match the user's language by default. The skill source and schema stay in English, and technical terms and formulas are kept.

Is it safe?

It runs two bundled Python scripts that read your paper file and write files into a work folder. SKILL.md forbids replacing or patching them in a normal run. PDF input needs PyMuPDF, which the script asks you to install with pip.

Information

updated Nov 2, 2026

Repository: yuan1z0825/nature-skills
Folder: skills/nature-paper-card

yuan1z0825/nature-skills skills.sh

Sources

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

Show all 5

nature-paper-card SKILL.md · github.com
nature-paper-card README_EN.md · github.com
nature-paper-card manifest.yaml · github.com
nature-skills README · github.com
prepare_paper.py · github.com