What does this skill do, and when should you use it?
This skill helps researchers produce submission-grade scientific figures for Nature-family and other high-impact venues, supporting both Python (matplotlib/seaborn) and R (ggplot2/patchwork/ComplexHeatmap) workflows. It guides the user through defining the core conclusion, evidence chain, data integrity, template compatibility, and export requirements before plotting, then loads the relevant backend quick-start and delivery preflight checklist. The skill includes multi-panel figure composition and journal-ready SVG/PDF/TIFF export, along with a separate OpenRouter GPT Image 2 route for AI-generated graphical abstracts and mechanism diagrams. It is part of an Apache-2.0 licensed monorepo and is installed as a standalone Agent Skill.
Reads manifest.yaml and static/core/contract.md and static/core/stance.md to load the figure contract and default operating stance, then resolves the plotting backend (Python or R) via a blocking gate that checks explicit user choice, language-specific inputs, or a saved preference in scripts/nature_figure_backend.py. Loads the corresponding backend fragment (static/fragments/backend/python.md or r.md) for the exclusive execution rule and publication quick-start. During figure building, it loads deep references on demand (figure-contract, asset-adaptation, template-catalog, api, r-workflow, design-theory, common-patterns, chart-types, nature-2026-observations, qa-contract, tutorials, demos). Before final delivery, it runs scripts/validate_figure.py on the plotting source and scripts/audit_pdf_text.py on the exported PDF, then requires a manual panel-by-panel audit. For AI-generated schematics, it reads references/openrouter-image-generation.md and optionally calls scripts/generate_openrouter_schematic.py to invoke the OpenRouter GPT Image 2 API.
- A PhD student needs to assemble experimental results into a multi-panel figure meeting Nature's layout standards; they describe the data and methods, and the skill guides the process.
- A researcher has existing matplotlib or ggplot2 code but needs it adjusted to journal requirements; the skill helps modify code and check export formats.
- A corresponding author needs to create a graphical abstract or mechanism diagram; they explicitly request OpenRouter GPT Image 2 generation and edit the draft output.
- An author responding to reviewer requests must audit and fix statistical labels, colors, and export quality in figures; the skill provides a QA checklist.
- A research team wants to standardize figure styling across publications; the skill serves as a template for consistent visual conventions.
How do you install this skill?
- OpenRouter external API call (GPT Image 2) may send data externally and depends on overseas service unreachable from mainland China; use cautiously and confirm user consent.
- Dependent validation scripts (validate_figure.py, audit_pdf_text.py etc.) not present in provided files; verify their availability and safety before use.
- Third-party figures4papers materials have unclear copyright; THIRD_PARTY_NOTICES.md retained but no usage granted, confirm copyright when using.
- Shell / CLI
- Local filesystem
PythonR
Install via npx skills: npx skills add Yuan1z0825/nature-skills --agent codex --skill nature-figure --yes --copy. For Claude Code, keep a stable clone and create a subagent or slash command pointing to skills/nature-figure/SKILL.md. Ensure Python or R environment with required plotting libraries is available.
tmp="$(mktemp -d)"
git clone --depth 1 https://github.com/Yuan1z0825/nature-skills.git "$tmp"
mkdir -p ~/.claude/skills
cp -R "$tmp/skills/nature-figure" ~/.claude/skills/
rm -rf "$tmp"Generated from the source repository and skill path; it copies only this skill's folder. If the author's install steps above differ, follow those first. To scope it to one project, replace ~/.claude/skills with that project's .claude/skills.
How do you use this skill?
Once installed, send your agent any of these to trigger it:
- Generate a submission-grade scientific figure based on this method and results.
- Use OpenRouter GPT Image 2 to generate a graphical abstract,
After installation, directly describe the task, e.g., 'Generate a submission-grade scientific figure based on this method and results.' The skill will ask whether to use Python or R and remember the default. For AI schematics, say 'Use OpenRouter GPT Image 2 to generate a graphical abstract,' and it will skip backend selection and call the API. Provide data files, method descriptions, or code snippets as input, and review the output for finalization.
What are this skill's strengths and limitations?
- Covers the full workflow from planning to delivery: contract definition, backend selection, template adaptation, and QA checks.
- Supports two main plotting languages (Python/R) with a memory mechanism for backend preference.
- Includes extensive deep references to handle varied figure complexity.
- Automated scripts validate the plotting source and exported PDF text, reducing manual errors.
- Clearly separates plotting and AI-schematic routes to avoid confusion.
- Relies on the user providing clear methods and data; vague descriptions may reduce effectiveness.
- Automated checks cannot replace manual panel-by-panel review.
- OpenRouter API route requires extra API key configuration, and output is treated as drafts requiring manual revision.
- The skill is specific to scientific figures and not suitable for interactive dashboards or pure data analysis.
How does this skill compare with similar options?
Side by side with related skills; every score comes from the same FSRS standard.
| Skill | FS score | Stars | Last updated | License |
|---|---|---|---|---|
| Nature Figure Making Skill (nature-figure) this page | 60 · Recommended | ★ 47k | 3d ago | Apache-2.0 |
| CSV Data Summarizer Skill | 45 · Not recommended | ★ 468 | 12mo ago | — |
| D3.js Visualization Skill | 60 · Recommended | ★ 231 | 11mo ago | — |
| Lieflat Charts | 57 · Use with care | ★ 6k | 1mo ago | NOASSERTION |
| TAO Standard Training Workflow ✓ NVIDIA · Official | 52 · Use with care | ★ 3.5k | 3d ago | Apache-2.0 |
How did FollowSkills review this skill?
Evidence shows the skill explicitly uses local Python/R backend and saves preference locally, no automatic external call; data integrity rules are clear and exclusions must be recorded, transparency is good. Third-party templates have copyright notice, no auto-grant. However, there is an external API call via OpenRouter (GPT Image 2), and no disclosure of network reachability, data sent, or user confirmation; script execution does not state least-privilege or rollback. No obvious malice but external call and permission details insufficient, hence deduction.
Instructions are internally consistent, flow clear (routing protocol, blocking gate, loading fragments). Dependent scripts (validate_figure.py, audit_pdf_text.py etc.) are not provided in the given files, so static review cannot verify their runnability; no specific failure feedback for abnormal inputs. Thus give 9, at static cap 10.
Target audience, use case (Nature submission figures) clearly described, Chinese support good, trigger words abundant (论文配图、科研绘图 etc.), but no explicit non-fit boundaries (though interactive dashboards excluded). Environment fit: depends on OpenRouter overseas service, likely unreachable from mainland China, no alternative provided, hence deduction.
README and SKILL.md structure well-organized, mirror in Chinese and English, with examples, boundaries, related skills. But missing version number, changelog, maintainer contact (except repository-level README), no known-limitation disclosure within skill (e.g., OpenRouter dependency), third-party materials copyright noted. Overall maintenance responsibility unclear, hence not full.
Description claims submission-grade figures, but static review sees no actual output verification or reproducible examples. Built-in templates and scripts may work, but lack direct evidence that output is directly usable or marginal value high. Static cap 7, give 6.
No tests, CI, or third-party independent verification covering nature-figure itself. Repository has CI but not for this skill. Project README claims DeepMind borrowed, but not verifiable. Static cap 5, give 4.
Open a dimension to read why it scored that way
Evidence confidence：Low — Mostly static review, author material or a limited demo; useful for discovery, not high-risk decisions.
See the full review method →