nature-skills for Claude Code and Codex: a skill pack for journal figures and statistics

nature-skills is a repository of figure, statistics, and writing skills aligned with the submission requirements of Nature journals. It holds 19 skills and a shared folder. This post covers the problem it solves, who it suits, how to install it and start a first run, and the one cost to check before you lean on it.

What problem it solves

The last stretch before submitting a paper is rarely the analysis itself. It is turning results into figures that meet the journal’s layout rules and making sure the statistical reporting is complete and consistent. Panels drift out of alignment, labels overlap, a p-value in the text disagrees with the table, and a multiple-comparison correction goes unmentioned. These are small errors, but reviewers notice them.

The pack splits that work by role. nature-figure creates Python or R plotting scripts from your data and includes scripts that check panel alignment and overlapping text. nature-statistics checks replicate counts, p-values, multiple comparisons, confidence interval notation, and numerical discrepancies between text and tables. Reference verification, academic English editing, and data availability statements each get their own skill, so you install only the part you need.

Who it fits

The chapter rates four situations. Three are enough to decide.

Situation	Fit	Reason
Turning analysis results into submission figures	Good fit	Provides scripts and checks
Checking statistical reporting before submission	Good fit	Checks numerical reporting items
Building internal BI dashboards	Poor fit	Focuses on submission requirements

Writing data availability statements sits in the middle, because that skill is still in draft status.

Install and first run

Codex can install a single skill. For Claude Code, clone the repository and copy it into the skills folder.

# Codex: install one skill you need
npx skills add Yuan1z0825/nature-skills --agent codex --skill nature-figure --yes --copy
# Claude Code: clone the repository and copy it to the skills folder
mkdir -p ~/ai-skills
git clone https://github.com/Yuan1z0825/nature-skills.git ~/ai-skills/nature-skills
~/ai-skills/nature-skills/scripts/autoupdate-skills.sh --force

The README says to use the frontmatter name shown by --list when installing with npx skills. Execution dependencies such as Python, R, browsers, and MCP are set up separately for each skill that needs them.

After installation, ask for the skill by name or describe the task in plain language.

[task to create submission figures or check statistical reporting]

Each top-level nature-* folder is one installation unit, and some of them need nature-shared alongside. The figure skill then produces plotting scripts from your inputs, and the statistics and data skills return checklists rather than invent sample sizes or DOIs. That last behavior is the point: the output tells you what is missing instead of filling the gap with a plausible number.

What trips people up

The catch is cost. Concept diagram drafts in nature-figure can call the OpenRouter API, and those calls are charged. Figure scripts may also pull in Python dependencies you do not have yet. If you only want plotting and layout checks, read which parts of the skill make external calls before you run it on a full figure set, and keep nature-statistics in its intended role as a check on reporting text, not a substitute for rerunning the analysis.

Where this comes from

This post adapts part of chapter 03-02 of The Claude Code & Codex Skills Guidebook. The full chapter covers the table of included skills with their stable and draft status, and the license caveat for the figures4papers examples bundled inside nature-figure, and the book has 52 more skills like this one. Read the full chapter in The Claude Code & Codex Skills Guidebook on Gumroad.