Nature Skills - An open-source AI research skills library, focused on the full paper workflow

What is Nature Skills

Nature Skills is an open-source research skills library for AI scholars worldwide, focused on the full Nature / CNS paper workflow. Nature Skills uses a routed SKILL.md architecture, with 18 built-in skills covering paper writing, polishing, peer review, figures, citations, literature search, and patent translation, among others. Nature Skills supports Codex、Claude Code one-click installation for multiple Agents such as these, with rules based on first-hand journal guidelines and outputs ready for direct use.

Main features of Nature Skills

• Paper writing: Draft Nature-style manuscripts by section and polish and translate them, supporting the full workflow from research plan to finished manuscript.
• Peer review and revision: Simulate Nature reviewers to output review reports, and automatically generate point-by-point responses and cover letters.
• Figures and statistics: Generate submission-grade research figures, and audit p-values, replicate numbers, and effect sizes in statistical reports.
• Literature management: Covers the full chain of search, citation, verification, download, and automated literature push.
• Close reading and reporting: Generate bilingual Chinese-English readers with paired text and images, structured close-reading cards, and presentation PPTs.
• Translation assistance: Handle data availability statements and experimental logging, and support converting papers into patent drafts.

How to use Nature Skills

• Install Node.js 18+: Ensure the local environment meets the runtime requirements.
• View available skills: Run npx skills add Yuan1z0825/nature-skills --list to list all skills.
• Install the skill package globally: Execute npx skills add Yuan1z0825/nature-skills --global --agent codex --skill '*' --yes --copy to deploy all skills to Codex.
• Start a new session: After installation is complete, restart the Agent session to ensure the skills take effect.
• Describe the research task: Tell the Agent directly in natural language what you want to do, such as "polish this abstract in Nature style".
• Automatic Agent routing: The system automatically matches and loads the corresponding skill rules based on the task content.
• Get ready-to-use outputs: Receive results such as text, figures, PPTs, or review responses output by the Agent.

Follow WeChat and reply "open source" to jointhe AI Open Source Project Community Group

Core advantages of Nature Skills

• Routed loading: Automatically matches and loads only the required rule fragments based on task, section, and journal, avoiding redundancy.
• Native to the Nature system: All rules are based on official Nature/CNS guidelines and published papers, not vague aesthetic preferences.
• Full-workflow closed loop: One-stop coverage from topic argumentation, writing and polishing, review simulation, to revision responses, figure creation, and literature management.
• Multi-Agent compatibility: Supports one-click installation and automatic updates for mainstream Agents such as Codex, Claude Code, and OpenClaw.
• Output is the product: Directly returns usable results such as paste-ready text, submission-grade figures, PPTs, and cover letters.
• Transparent and explainable rules: Each writing/review rule states its source and rationale, explicitly marking missing evidence and applicable boundaries.
• International influence: The design philosophy has been adopted by Google DeepMind, which launched Science Skills, gaining recognition from top institutions.
• Modular and extensible: Each skill is self-contained; new features can be added with zero intrusion into the existing system, and the community can contribute freely.

Project address of Nature Skills

• GitHub repository：https://github.com/Yuan1z0825/nature-skills

Comparison of Nature Skills with similar competing products

Application scenarios of Nature Skills

• Drafting first drafts of papers for the Nature system: Based on research data and figures, automatically generate sections such as abstract, introduction, methods, and discussion in Nature style, and reconstruct the argumentation logic.
• Pre-submission peer review simulation: Before formal submission, simulate three mutually blind Nature reviewers to output review reports, identify manuscript weaknesses in advance, and classify comments as Major/Minor.
• Point-by-point response to revision comments: Parse journal revision emails, independently generate point-by-point responses, a cover letter, and highlighted positions in the revised manuscript for each reviewer.
• Submission-grade scientific figures and statistical review: Generate Python/R scientific figures that meet Nature requirements, and review compliance of p-values, replicate numbers, effect sizes, etc. in statistical reports.
• Full-chain literature management and intensive reading reports: Automate retrieval, verify citations, generate bilingual Chinese-English readers, and convert papers into Chinese group meeting presentation PPTs with one click.