lwq-star/ultraplot-figures

UltraPlot Figures is a skill for the Codex AI assistant that helps produce, edit, and check static scientific figures using a Python plotting library called UltraPlot. It only works inside a Codex Desktop or Codex CLI conversation that supports loading local skills. It is not a plotting program you install and import into your own Python code. Once installed, you tell Codex what data you have, what the figure should show, and what output files you need. Codex then writes an editable Python script, produces the figure, and checks that the exported files meet common requirements. It can also revise a plotting script you already have or inspect an existing PDF, TIFF, SVG, or PNG figure. The project exists because general AI assistants can write plotting code quickly, but code that runs is not the same as code that is reproducible, scientifically clear, or ready to submit to a journal. This skill adds a workflow on top of that: it suggests a suitable figure type based on your data and audience, produces runnable code that separates heavy data processing from the plotting itself, handles common needs like multi panel layouts, journal specific sizing, colorblind friendly colors, legends, maps, and specialized scientific chart types, and checks the finished output against your original request. To install it, you ask Codex to fetch the skill directly from its GitHub repository, or you download it yourself and place it in Codex's skills folder. You will also need Python 3.10 or newer with UltraPlot, Matplotlib, and NumPy available, plus a few extra libraries depending on the file types you are working with, such as Pillow for images or openpyxl for Excel files. The authors are explicit that any figure this skill produces still needs a human author to review it before it goes anywhere near a publication. It can check formatting and file settings, but it cannot judge whether the underlying data, methods, or conclusions are correct.