gitmyhub

nature-skills

Python ★ 19k updated 2d ago

符合nature论文学术表达和科研绘图的Skill

Collection of nine reusable skill bundles for AI coding assistants that help write scientific papers in the style of Nature journals, covering figures, prose polishing, citations, and rebuttals.

PythonRMarkdownClaude CodeCodexsetup: easycomplexity 2/5

This repo is a collection of reusable instruction bundles, called skills, that help AI coding assistants do tasks specific to writing scientific papers in the style of journals like Nature. It is maintained by a PhD student at Shanghai Jiao Tong University whose own work is in medical AI. Each skill lives in its own folder under skills/ and includes a SKILL.md file plus supporting reference materials, scripts, and example assets. The folder is the unit of installation, not just the SKILL.md file. The README lists nine skills. nature-figure builds multi-panel publication-style figures in Python (or R) following journal visual standards, with editable SVG output. nature-polishing rewrites academic prose to match Nature's house style, and nature-writing drafts manuscript sections like abstracts and introductions. nature-citation retrieves and formats citations for Nature-family journals and can export to ENW, RIS, or Zotero RDF. nature-data helps write Data Availability statements and check FAIR data practices. nature-reader turns a paper into a bilingual Markdown reader that keeps figures and source anchors aligned. nature-response builds point-by-point reviewer rebuttal letters, nature-paper2ppt converts a paper into a Chinese PPTX presentation, and nature-academic-search runs multi-source literature search with DOI verification. Installation depends on which assistant you use. For Codex, the README tells you to clone the repo and copy the skills/nature-* folders into ~/.codex/skills/, then restart Codex. For Claude Code, the recommended path is the plugin marketplace: run /plugin marketplace add with this repo's URL, then /plugin install nature-skills and /reload-plugins, after which all nine skills are available. There is also a manual subagent option for users who want to enable skills one by one. For other agents, the README suggests copying the whole skill directory into your prompt library and adapting the frontmatter to that agent's format. The project welcomes issues and pull requests, has a WeChat support group, and is recruiting medical-AI interns.

Where it fits