Copyplot

CopyPlot

科研图片一比一复刻，只需一张截图，即可得到代码+数据，让你也能画出Nature级的精美科研图片！

CopyPlot is a Codex skill for reverse-engineering scientific figure images into editable, reproducible plotting projects. 给它一张论文中的科研图，CopyPlot 会识别图形类型与视觉参数，生成近似示例数据、Excel 工作簿、可运行的 Python 绘图代码，并通过渲染差异进行迭代检查。

效果示例

目标图片：

Integrated global assessment of the natural forest carbon potential. Nature 624, 92–101 (2023). https://doi.org/10.1038/s41586-023-06723-z

提取数据：

复刻结果：

注：复刻数据不代表原论文数据，仅为重现绘图风格。

Features

• Classify line, scatter, bar, histogram, box, violin, heatmap, contour, error-bar, multi-panel, and mixed figures.
• Infer axes, scales, colors, line widths, markers, legends, typography, margins, and annotations.
• Recover defensible digitized approximations or create clearly labeled synthetic look-alike data.
• Store example data in Excel whenever practical.
• Generate runnable Matplotlib/Seaborn code plus PNG and vector output.
• Compare the recreation with the source using pixel and edge diagnostics.
• Clearly distinguish inferred data from original scientific measurements.

Install

Ask Codex:

Install the CopyPlot skill from https://github.com/YishuoHao/copyplot/tree/main/skills/copyplot

Or install with the bundled skill installer:

python ~/.codex/skills/.system/skill-installer/scripts/install-skill-from-github.py \
  --repo YishuoHao/copyplot \
  --path skills/copyplot

Restart Codex or begin a new turn after installation.

Usage

Attach a scientific figure and prompt:

Use $copyplot to analyze and recreate this scientific figure with runnable Python code and Excel example data.

中文示例：

使用 $copyplot 分析并复刻这张科研图，生成 Python 代码和 Excel 示例数据。

Output

A typical result contains:

recreated-figure/
├── recreate_plot.py
├── example_data.xlsx
├── recreated_plot.png
├── recreated_plot.pdf
├── analysis.json
└── qa/
    ├── comparison.json
    └── difference.png

Raster figures do not contain their original numerical datasets or plotting commands. CopyPlot labels inferred and synthetic values as approximations and should not be used to claim recovery of source-paper data.

Repository structure

The installable skill is located at skills/copyplot.

License

MIT