scientific-plotting-skill
dazhiyang · dazhiyang/scientific-plotting-skill
This AI skill helps coding agents generate publication-style scientific plots using ggplot2 or plotnine, with specific formatting for vector PDF export and color palettes.
AI skill: publication-ready ggplot2 / plotnine figures, vector PDF, Wong + viridis.
Preview
What you can do with it
- Create journal-ready figures for a research paper from a CSV dataset.
- Produce vector PDF plots with specific dimensions and typography for scientific presentations.
- Ensure colorblind-friendly continuous and discrete color schemes in data visualizations.
Safety check
No risky patterns found in the files we scanned.
Automated scan of the repo files. Always review code you run.
From the README
📊 Scientific Publication Plotter (AI Skill)
Overview
This repo ships one skill file, SKILL.md. Point your coding agent at it when you want publication-style figures: ggplot2 in R or plotnine in Python, vector PDF, journal-friendly widths (85 mm / 180 mm), Wong discrete colors, viridis with quantile splits for continuous color, and typography rules (single font size, no plot title—caption in the manuscript instead).
No bundled scripts: the agent generates code in your project from the rules in SKILL.md.
Features
| 📏 Sizing | Default 85 mm single column, 180 mm when you need double column; prefer flat figures unless x and y are the same quantity. |
| 🔤 Typography | Times serif, one point size for all figure text; no plot title or subtitle on the canvas. |
| 📄 Export | Vector PDF (or EPS); raster only as an optional preview. |
| 🎨 Color | Wong order for discrete (≤8); viridis family + equal-count quantiles for continuous fields. |
| 🌐 Stack | ggplot2 (R) and plotnine (Python) only for the figure code. |
| 🗺️ Heavy plots | Guidance for dense scatter (e.g. scattermore in R) and low-resolution map outlines so PDFs stay light. |
Example prompts
Use wording close to your real task so the agent picks up the skill:
- “Read
SKILL.mdand plot this CSV for my manuscript—single column PDF, ggplot2.” - “Publication-style plotnine figure, Wong colors, vector output.”
- “Journal figure: 85 mm wide, no title, Times, save PDF.”
How to use
🖱️ Cursor
- 📋 Copy
SKILL.mdinto your.cursor/rules/folder. - ✏️ Rename it to
SKILL.mdc. - 💬 Ask the AI to follow the skill (e.g. “Plot my data for the manuscript.”).
🤖 Claude Code, Copilot, and other agents
- 📁 Add
SKILL.mdat the root of your project (or path your agent reads). - 💬 Ask the model to use it (e.g. “Read
SKILL.mdand plot the attached data.”).
🛸 Google Antigravity
Antigravity loads Agent Skills from a folder that contains SKILL.md (see Google’s Authoring Antigravity Skills codelab). This repo’s SKILL.md includes YAML frontmatter (name, description) so the agent can match plotting tasks.
Project-only (recommended)
- 📂 Create e.g.
.agent/skills/scientific-publication-plotter/in your repo. - 📋 Copy
SKILL.mdthere (path must end withSKILL.md). - 💬 Ask for a manuscript-style figure in chat; the skill should load when the request matches the description.
All projects on this machine
- 📂 Create
~/.gemini/antigravity/skills/<your-skill-name>/and placeSKILL.mdinside.
Optional — rules UI: Agent Manager → ⋮ → Additional options → Customizations for GEMINI.md / AGENTS.md (Antigravity rules). Pasting the body of SKILL.md there is possible; .agent/skills/… keeps the skill on-demand.
License
MIT — see LICENSE.