EditaPlot: an AI skill that drives local Origin to make editable figures

What EditaPlot automates, and who it is for

EditaPlot is a Codex skill for Windows that turns experimental data into publication figures by driving Origin, the commercial plotting software, rather than by drawing them in Python. You hand it a data file, and it reads the table, explains what each column is for, recommends one to three chart types with colors, asks you to confirm the figure elements, then calls Origin and verifies the result. The output that matters is an editable OPJU project, exported alongside PNG, PDF and TIF. The audience is researchers who already work in Origin and want an assistant to build the figure while keeping the scientific choices in their own hands. It is aimed squarely at people on Windows with an Origin license, not at anyone looking for a standalone image generator.

From columns to an editable Origin project

The mechanism is a guided pipeline with a confirmation gate. Before plotting, EditaPlot sorts every column into a role and asks you to confirm it in plain language: primary evidence to be drawn, visible auxiliary such as a fit line or reference marker, compute-only data kept for weighting or layout, retained but not drawn, or still uncertain. A column marked uncertain stops the planning rather than being guessed into a new curve. Once you confirm the scientific purpose, it plots inside a dedicated Origin instance it launches itself, then reads the objects back and checks that the raw data was not altered and that axes and text are intact. If the source file or the column mapping changes, the earlier confirmation is invalidated and has to be redone, which keeps the figure tied to what you actually approved.

Running it on Windows with a local Origin

EditaPlot runs on Windows 10 or 11 x64 with Python 3.10 to 3.12, and it targets Origin or OriginPro 2021 through 2026b, though the README states it is fully verified only against Origin 2024b. You do not open Origin yourself; the skill launches a dedicated instance before plotting and will not install or modify your Origin. The repository ships a Windows launcher:

editaplot.cmd

With the skill installed, you drive it from Codex by naming the $editaplot trigger and describing what you want, for example asking it to use a specific palette while keeping all data. It then walks you through the column roles and element list before it draws anything, and hands back the OPJU and the exported images once its checks pass.

The reference-image route and its hard limits

You can also upload a PNG, JPEG or TIFF as a style reference, and this is where the project's discipline shows. EditaPlot treats a reference image as a graphic brief: it extracts the panel layout, the point, line and bar elements, the data encoding and a limited visual style, then lists only the parts that both fit the current template and are backed by your own data for you to confirm. It explicitly does not reverse-engineer experimental values from pixels, does not copy numbers, text, fit results, phases, logos or watermarks, and does not embed the reference picture into the OPJU. Where a key element cannot be expressed by the current template, it says so and keeps the template default rather than pretending support exists. Your explicit choices, such as exact series colors or line widths frozen into a visual-style JSON, take priority over the reference.

Where it will not go: Windows-only, Origin-required, no invented analysis

The limits are stated plainly and matter for adoption. Version 1 is verified only on physical Windows 10/11 x64 machines, so there is no macOS, Linux, WSL, Wine, or virtual-machine path; a Mac user cannot run the full Origin flow today. It requires a licensed local Origin, which is commercial software, so this is not a free end-to-end tool despite the skill itself being open source. And it deliberately refuses to do analysis for you: it will not smooth data, remove outliers, add peaks, compute errors, fit curves, identify phases, calculate a band gap or Rwp, or compute SHAP values. Results such as lifetimes or SHAP contributions are only drawn when you provide them, so EditaPlot is a figure builder, not an analysis engine.

EditaPlot versus a matplotlib-based plotting assistant

The natural alternative is an AI plotting skill that generates matplotlib or seaborn code and renders an image or a script. The difference in approach is concrete. Such a tool produces a static picture or Python source, which is portable and needs no commercial software, but the result is not an Origin project and cannot be reopened and refined in the workflow many labs already standardize on. EditaPlot instead automates Origin itself and hands back a true editable OPJU, at the cost of requiring Windows and an Origin license. The README is also explicit that it will not let a Python preview stand in for an Origin figure. So the choice is about the deliverable: reach for a matplotlib assistant when you want a free, portable image, and for EditaPlot when the endpoint has to be an editable Origin project.

Apache-2.0, verification status, and domain coverage

EditaPlot is released under Apache-2.0, a permissive license with a patent grant, so the skill code is free to study and reuse; the Origin dependency and your data remain yours to license and provide. On maturity, there are no tagged releases yet, the last push was on 2026-09-17, and the honest verification claim is narrow: the compatibility target is Origin 2021 to 2026b, but full verification is reported only on 2024b, so treat other versions as untested. The domain coverage is wide, spanning materials and spectroscopy figures such as XPS, XRD Rietveld, FTIR, NMR and UV-Vis, general statistics, distribution plots, and medical or deep-learning figures including ROC, calibration and SHAP dashboards, with example CSV files committed so you can learn each expected format first.

Editorial conclusion

Adopt EditaPlot if you are on Windows 10/11 x64, hold an Origin license, and want an assistant that builds an editable OPJU while leaving the scientific decisions to you. Do not expect it on macOS or Linux, and do not expect it to fit, smooth or compute anything you have not supplied. Before relying on it, confirm your Origin version, since full verification is reported only on 2024b, and start from the committed example CSV files to learn the column formats it expects.

Frequently asked questions

Official sources

1. hang-jin/editaplot on GitHub
2. Issues
3. License: Apache-2.0
4. README