What it does

nature-skills is a bundle of reusable prompt workflows—“skills” in Codex parlance—that automate the tedious scaffolding of high-impact academic publishing. Each nature-* directory is a self-contained instruction set: generate Nature-compliant figures, polish prose to journal style, draft reviewer responses, build bilingual paper readers, or convert manuscripts into Chinese journal-club slides. Install via Codex plugin marketplace, Claude Code subagents, or manual copy-paste.

The interesting bit

The project treats prompt engineering as infrastructure. Rather than one-off ChatGPT queries, it version-controls the process: SKILL.md files encode workflows, references/ hold style guides, and static/ bundles assets so the LLM doesn’t hallucinate figure palettes or citation formats. It’s essentially CI/CD for academic writing—if your CI pipeline had opinions about serif fonts and pan-cancer cohorts.

Key highlights

• 10 specialized skills covering figures, writing, polishing, peer review simulation, citation management, data availability statements, and paper-to-PPT conversion
• nature-figure generates multi-panel matplotlib plots with Nature typography, semantic color palettes, and editable SVG output; includes 5 gallery examples and a 10-chart-type atlas
• Bilingual support: nature-reader produces source-anchored, figure-grounded Markdown translations with original text side-by-side
• Codex-native: packaged as a plugin marketplace bundle with shared dependency handling via skills/_shared/
• Claude Code compatible: thin subagent wrappers point to cloned repo so supporting files stay accessible
• Citation rigor: nature-citation exports ENW, RIS, and Zotero RDF with strict Nature/CNS-family formatting

Caveats

• Several skills marked Draft or Beta (nature-writing, nature-reviewer, nature-data, nature-reader, nature-response, nature-paper2ppt, nature-academic-search); stability varies
• README is upfront that the repo doubles as personal networking for the author’s medical-AI startup and recruiting for their Shanghai Jiao Tong lab—useful context, slightly unusual for a tool project
• Manual installation requires careful path handling; _shared/ must stay adjacent to skill folders or references break

Verdict

Worth a look if you’re a grad student or postdoc drowning in formatting guidelines and want to automate the mechanical parts of publishing. Skip if you need battle-tested, fully stable tools today—the draft-status skills are promising but uneven.