Yuan1z0825/nature-skills

符合nature论文学术表达和科研绘图的Skill

nature-skills — AI-agent skills for scientific research

nature-skills is a library of reusable, packaged workflows (called "skills") that extend AI coding agents to do real academic research work. Each skill is a folder containing a SKILL.md instruction file plus supporting scripts, templates, and reference materials. You install them into agents like Claude Code, Codex, or Chatbox, then describe a task in plain language — the agent reads the skill's instructions and follows its workflow.

What it covers

The repo ships 22 skills spanning the research lifecycle:

  • Reading & understanding: nature-reader (bilingual Markdown paper reader with figures/formulas), nature-paper-xray (deep reasoning reconstruction), nature-paper-card (structured paper analysis)
  • Writing & polishing: nature-polishing (Nature-style English rewriting), nature-writing (drafting manuscript sections), nature-statistics (statistical reporting review)
  • Figures & presentations: nature-figure (publication-grade plots in Python/R), nature-paper2ppt (paper → Chinese PPTX), nature-image2ppt (rebuild editable PowerPoint from images)
  • Review & revision: nature-reviewer (simulated blind peer review), nature-response (drafting rebuttals and cover letters)
  • Literature & citations: nature-academic-search (multi-source search, citation verification), nature-citation (CNS-restricted reference export), nature-ref-verifier (cross-checking reference metadata), nature-downloader (legal full-text retrieval), nature-literature-pipeline (automated daily literature discovery)
  • Other: nature-paper-to-patent (Chinese patent drafting), nature-data (Data Availability statements), nature-experiment-log (Obsidian experiment logging)

A shared package, nature-shared, provides common support files and must be installed alongside skills that depend on it.

How installation works

  • npx skills (Node.js ≥18): install selected skills into Codex or other agents, e.g. npx skills add Yuan1z0825/nature-skills --agent codex --skill nature-figure --yes --copy
  • Claude Code: clone the repo and create lightweight subagent or slash-command wrappers pointing at the real SKILL.md files; optional SessionStart hook auto-updates via scripts/autoupdate-skills.sh
  • Codex: run scripts/update-codex-skills.sh --pull to sync all skills, with --check verification and --prune cleanup
  • Chatbox: GUI-based "Skills" settings let you scan the GitHub repo and install selected skills
  • Other agents (OpenClaw, OpenCode, Hermes): documented in docs/open-source-agent-frameworks.md

The auto-update script is safety-conscious: it throttles network checks (default 1/hour), skips silently on failure, refuses to force-pull over uncommitted changes, and logs per-target.

Notable details

  • Skills are in different maturity states: Stable (e.g., nature-figure, nature-ref-verifier), Beta (most), and Draft
  • Some skills need extra runtime dependencies (Python packages, Playwright for browser automation, MCP servers with API credentials)
  • A web version at natureskills.cn exposes the core polishing and figure features
  • The README also promotes the founder's paid services (GPT/Claude top-up store, paid knowledge community) and claims Google DeepMind's "Science Skills" was inspired by this project's design

Bottom line

A practical, actively maintained skill library that turns AI coding agents into research assistants — install a skill, describe a task, and the agent executes a reproducible, source-grounded workflow. The heavy promotion of commercial services in the README is a caveat, but the skills themselves are substantive and well-documented.

Related

  • Project
    LeonChaoX/qinyan-academic-skills

    A library of 187 installable AI agent skills for academic research, featuring a specialized suite for Nature-style scientific writing and manuscript preparation.

  • Project
    Boom5426/Nature-Paper-Skills

    A collection of 27 AI agent skills for Codex and Claude Code to help scientists write, revise, and manage rebuttals for Nature-series life science and computational biology papers.

  • Project
    jamditis/claude-skills-journalism

    A library of Claude Code (and limited Codex) plugins/skills that give AI agents journalism‑specific capabilities—AP‑style enforcement, fact‑checking, FOIA templates, PDF design, video analysis, research tools, etc.—installable via Claude’s marketplace or by copying `SKILL.md` folders.

  • Project
    xstongxue/best-skills

    A collection of high-quality SKILL.md files for AI agents (Cursor, Claude Code, etc.) that enable automatic task triggering for academic writing, diagramming, and development workflows.

  • Project
    claesbackman/AI-research-feedback

    A collection of Claude Code skills that provide multi-agent academic reviews for research papers, grant proposals, and pre-analysis plans, including paper-code reproducibility checks.