Guide · Agent skill
How to use the Nature Paper Reader agent skill (nature-reader)
Updated · written with AI from the project's own documentation and checked against our catalogue data
nature-reader is an agent skill that tells an AI agent how to work through a scientific paper and produce a Chinese-English reading document. It is for students, researchers and translators who want a bilingual reading of a paper, source-linked answers to specific questions, or consistent terminology across a document. After this guide you will know what the skill does, how to install it, how to phrase requests so it triggers, and what to check in the output.
Nature Paper ReaderYuan1z0825/nature-skills · Apache-2.0 Download the skill Quick start
- 1Download the ZIP. Download the skills/nature-reader ZIP from this directory's source page for Yuan1z0825/nature-skills.
- 2Unzip into a skills folder. Unzip the folder into ~/.claude/skills/ for personal use, or into a project's .claude/skills/ folder for one project.
- 3Or upload in the Claude apps. In the Claude apps, open Settings → Capabilities → Skills and upload the skill there.
- 4Confirm Python 3 is available. Make sure Python 3 is installed so the agent can run scripts/validate_reader_math.py.
- 5Provide the source paper. Give the paper as a PDF, HTML page, DOI or arXiv link, or pasted text.
- 6Make a natural request. Ask for a full translation, a bilingual excerpt, a close reading, or a source-linked question, and name the scope you want.
- 7Check the outputs. Read translation_notes.md first, then compare paper.md and source_map.json against the source for anything you will cite.
What is an agent skill, and what does nature-reader make the agent do?
An agent skill is a folder of instructions and optional scripts that an AI agent loads when a task matches the skill's description. nature-reader covers paper reading: it converts a paper PDF, DOI, arXiv link, publisher HTML or pasted text into a Markdown reader with Chinese and English text side by side.
The skill keeps figures, tables and equations tied to their location in the source. It also builds a glossary of recurring terms so the same term is translated the same way throughout. For follow-up questions, it answers with citations back to the source. If you ask for only an excerpt or a single question, it works on that part and does not generate a full reader.
- Full bilingual reader: paper.md with aligned Chinese and English text.
- Source map: source_map.json links paragraphs, figures and tables to page and source anchors.
- Translation notes: translation_notes.md records terminology, unclear content and translation decisions.
- Equations: rendered as Markdown math, each display equation given a stable E anchor.
- Figures: cropped images saved under assets/, with translated legends placed near the first substantive discussion.
Install the nature-reader skill
Download the ZIP of skills/nature-reader from the Yuan1z0825/nature-skills source page on this directory. Unzip it into one of two locations, depending on how you want to use it.
For a personal install that works in any project, unzip it into ~/.claude/skills/. For a single project, unzip it into that project's .claude/skills/ folder. If you use the Claude apps instead of a local setup, upload the skill under Settings → Capabilities → Skills.
- Personal (all projects): ~/.claude/skills/nature-reader
- Single project: .claude/skills/nature-reader inside the project folder
- Claude apps: Settings → Capabilities → Skills, then upload the skill
Check what you need before you start
The skill needs three things: an agent that can load skills, a source paper, and Python 3 for the bundled checker script. The source can be a selectable-text PDF, a publisher or preprint HTML page, a DOI or arXiv link, or pasted text.
The skill runs scripts and uses the network. The validation script is scripts/validate_reader_math.py, and the skill may fetch a paper when you give it a DOI or arXiv link. Review the files before installing if you are working with confidential material.
- Python 3 (to run scripts/validate_reader_math.py)
- A source paper: PDF, HTML, DOI/arXiv link, or pasted text
- Network access, because the skill runs scripts and may resolve DOI or arXiv links
- Licence: Apache-2.0
Trigger the skill with natural requests
You do not need a special command. Ask in plain language and include the paper. The skill's description lists trigger phrases such as 全文翻译, 中英文对照 and 论文精读, and also covers source-linked questions about a paper.
To get a full reader, ask for one. To get a narrow answer, name the excerpt or question. The skill is designed to respect a requested excerpt and not expand it into a full reader.
- 全文翻译：把这篇论文做成中英文全文对照。 (full-text translation)
- 中英文对照：翻译第3节并保留公式编号。 (bilingual excerpt)
- 论文精读：精读这篇论文，图表放在对应正文附近。 (close reading)
- Source-linked question: 根据这篇论文，作者如何定义该指标？请给出出处。 (question with citation)
- Excerpt only: 只翻译摘要和引言，不要生成完整阅读材料。 (limited scope)
Review the output
A full run produces a folder with the reader and its supporting files. Start with translation_notes.md to see what the agent could not confirm, such as missing pages, low-confidence figure crops or formulas it could not transcribe reliably.
For formulas, the skill keeps the original image crop when it cannot transcribe an expression with confidence, rather than substituting a guessed LaTeX version. Check paper.md against the source for any equation you plan to cite.
- paper.md: the full reader
- source_map.json: page, text block and figure/table source mapping
- translation_notes.md: terminology, gaps and notes
- assets/: figures, cropped images and attachments
- assets/equations/: original equation crops, created when needed
What people use it for
- Turning a paper into a bilingual Chinese-English reading for coursework or a reading group.
- Asking specific source-linked questions about a method, dataset or result, with citations back to the paper.
- Translating one section or abstract without producing a full reader.
- Keeping terminology consistent across a multi-paper translation project using the glossary.
Frequently asked questions
How do I install the nature-reader skill?
Download the skills/nature-reader ZIP from this directory's source page. Unzip it into ~/.claude/skills/ for personal use or into a project's .claude/skills/ folder, or upload it in the Claude apps under Settings → Capabilities → Skills.
What does nature-reader need to run?
It needs Python 3 for the bundled script scripts/validate_reader_math.py and a source paper as a PDF, HTML page, DOI or arXiv link, or pasted text. It also uses the network, and it runs scripts.
Is nature-reader free?
The skill is published under the Apache-2.0 licence. Check the licence file in the download for the terms that apply to your use.
Why does the skill not produce a short summary?
The skill is designed to produce full readers and deliberately avoids summary mode. For a quick overview, a summary-focused tool or a direct request to the agent will be faster.
Can it handle scanned PDFs?
Not reliably without OCR. Its accuracy depends on extractable source material, so scanned or image-only papers without OCR support may need manual proofreading.
Are there alternatives or related skills?
The skill's README lists related skills in the same family: nature-downloader for fetching full text legally, nature-paper2ppt for turning a reading into a Chinese presentation, and nature-citation for extracting citations. Use a general translation tool if you only need a quick, unstructured translation.
New agent skills, repos and AI tools — every week
The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.
Our verdict
Use nature-reader if you need a structured Chinese-English reading of a paper, source-linked answers to specific questions, or consistent terminology across a translation project. Pick something else if you only need a quick summary, if you plan to publish full translations of copyrighted papers, or if your papers are scanned images without OCR.
Download the skill