EditaPlot

Give Codex the table. Get an editable scientific figure back in Origin.

Understand the data · Choose the view · Customize the style · Verify the deliverables

Start · Gallery · Prompts · Guide · 中文

I built EditaPlot to spend less time repeatedly tuning figures and more time discussing the data itself.

Tell Codex what the table contains and what you want to show. It explains each column first, waits for confirmation, then uses local Origin and checks the result.

EditaPlot is a Codex Skill, not a standalone plotting application. You can continue editing curves, axes, and legends in Origin. Within each verified route, you can also specify supported colors, line widths, and layouts.

Requirements: a physical Windows 10/11 x64 computer, Codex, and local Origin/OriginPro. You do not need to open Origin first. macOS, Linux, WSL, and virtual machines are currently unsupported.

See the results

XPS observations and supplied components Display available backgrounds, envelopes and components without inventing missing peaks.	Grouped boxes and raw observations Show each observation alongside the group distribution.
SHAP importance, contribution, and beeswarm Use upstream per-sample SHAP values; the plotting layer does not train a model.	30 × 30 dense heatmap Keep all 900 source values and row/column order, without interpolation or automatic clustering.

All examples use synthetic teaching data and real Origin output. They are not experimental or clinical findings. Click an image for the full-resolution file.

Explore all 45 plotting routes →　Fluorescence and EEG signals　中文图库

Get started

1 / Install: give this request to Codex

You do not need to learn Git first. Copy this into Codex:

Please install EditaPlot from the latest main branch:
Repository: https://github.com/hang-jin/editaplot
Keep the complete repository, including runtime, and run its root editaplot.cmd setup and doctor.
Reuse a compatible 64-bit Python 3.11/3.12 when available; ask before installing Python.
Do not install or modify Origin, change the registry/DCOM, request administrator rights, or use mouse control.
Report whether the environment is ready, then ask me for the dataset.

Alternatively, download the complete ZIP, extract it, and let Codex install from that folder.

Do not copy only skill/editaplot. It needs the bundled drawing engine.

git clone https://github.com/hang-jin/editaplot.git
Set-Location editaplot
.\editaplot.cmd setup
.\editaplot.cmd doctor

2 / Provide the data: say what you want to compare

Start a new Codex conversation, invoke $editaplot, and attach CSV, TXT, XLS, or XLSX data. For example:

Use $editaplot to inspect this dataset. Do not draw yet and do not modify the source file.
Explain each column in plain language, recommend no more than three suitable charts, and tell me what still needs confirmation.

Chinese headers and file paths are supported. Ask for suggestions if you are unsure, or name a chart such as grouped boxes, XPS comparison, or precomputed SHAP.

If the headers need reorganizing, open the matching template example or ask Codex to compare your table with it.

Helper columns are created only in memory or the editable Origin workbook. The source file remains unchanged.

3 / Confirm and draw: receive files you can keep editing

After confirming groups, units, uncertainty meaning, and visible figure elements, continue with:

Use the plan we just confirmed and draw it through Origin. Export an editable OPJU plus PNG, PDF, and TIF.
Check the data mapping, axes, fonts, and exported appearance, then briefly report the output location and verification result.

I have the Skill verify the local drawing path before rendering the formal dataset. It does not invent measurements, fitted components, or uncertainty values missing from the source.

By default, results stay together in a new folder beside the original dataset:

my-data.csv
my-data_EditaPlot_<time>/
  render-plan.json
  result.opju
  result.png
  result.pdf
  result.tif
  …readback and verification reports

Add “keep the editable Origin window open” if you want to continue adjusting it.

A picture alone is not completion. A formal delivery also needs the real OPJU, other exports, and data and figure checks. See the full guide.

Three common requests to copy

I am not sure what to plot

Use $editaplot to inspect this table read-only. I want to compare differences among groups, but I have not chosen a chart.
Recommend no more than three options, explain the required columns, and wait for my confirmation before drawing.

I already know the chart I want

Use $editaplot to draw grouped boxplots with the raw observations overlaid.
First confirm the groups, Y-axis unit, meaning of each observation, and sample sizes. Do not add significance marks automatically.

I want to follow a reference figure's style

Use $editaplot to take color, line-width, and layout cues from this reference. Use only my table for data.
First list what the current template can adopt, what must keep its default, and what cannot be implemented. Wait for my confirmation before drawing.
Do not copy the reference's data, conclusions, sample names, or watermark.

Choose by your data

I currently maintain 45 public plotting routes. One route can accommodate different series counts, headers, and supported display modes.

You do not need a separate template for every material. Inspect the input example before using your own data.

XPS, GSAS, SHAP, lifetime, and bandgap workflows have specific scientific boundaries.

The drawing layer does not silently fit, identify peaks, train models, reconstruct images, or calculate spectra from raw EEG. Required upstream results and unresolved choices are stated first.

Capabilities depend on the route and local checks.

Choose your reference style and colors

Choose among template defaults, reference-guided suggestions, and exact custom values. Your explicit choices take precedence.

A reference provides compatible grammar and style, not borrowed values, conclusions, sample names, or watermarks.

For XPS, for example: “Raw #173F5F, 2.4 pt lines, 38% fill transparency, an 18 × 18 cm page, and no legend.”

Supported adjustments differ by template. Codex reports each request as applied, default retained, or rejected. It does not present an unsupported parameter as if it worked.

Chinese palette selector · Reference and exact-style policy

What you need to know before using it

Grant only task-scoped access. Read the repository and selected data. Write only to the project, the current user's $HOME.codex\skills\editaplot, and the data folder.

Allow local commands and a dedicated Origin instance. Network access is needed for downloads and dependency updates.

Normal use needs no administrator rights, mouse control, whole-drive write access, or DCOM, registry, firewall, or Origin-installation changes.

For a protected output directory, permit only the current folder or explicitly choose another writable destination.

Deidentify sensitive data first, including names or IDs left in reference images. I do not claim that the Skill finds every sensitive detail.

The local plotting runtime does not initiate a network upload. Files given to Codex remain governed by your account, organization, and retention policies. See Privacy.

Go deeper

Star trend

If EditaPlot saved you time, a Star helps other researchers find it. I record only daily repository totals—not account lists, IDs or personal star timestamps.

Open source, contributing, and support

I independently maintain EditaPlot under Apache License 2.0. It is not affiliated with, sponsored by or endorsed by OriginLab Corporation; product names describe compatibility only.

Useful feedback includes a shareable minimal example and clear reproduction steps. Do not put medical records, private data, credentials or full private paths in Issues.

Contributing · Support scope · Security reports · Privacy

Buy me a coffee ☕

If EditaPlot helped, you are welcome to buy me a coffee. Even a small tip encourages continued work on templates and tutorials.

Tips are entirely optional. They do not unlock features, change support priority or affect Apache-2.0 rights. Sharing the project or reporting a useful issue is just as welcome.