Overview
This is an Agent Skill -; a folder with a SKILL.md file that Claude
reads when a data-visualization request comes in. You describe the chart you want
("a bar chart comparing average test scores across grade levels from this CSV"),
and Claude inspects the data, picks an appropriate chart type, reshapes it if needed,
applies sensible styling, and saves the result as a file. You never have to mention
"Seaborn" or any chart type, and you never write the Python.
It's aimed at the moment where you've pulled data into Excel, a CSV, or a pandas DataFrame and want clean charts out the other end without the matplotlib-and- Stack-Overflow detour.
How It Works
The skill teaches Claude a consistent workflow: inspect the data first, choose a
chart type, reshape if needed, plot, label, save, summarize. The single most
important thing it encodes is the long ("tidy") versus wide data distinction and
how to reshape between them with pd.melt -; the thing that trips
people up most often with Seaborn.
Beyond that it carries a decision table mapping intent to chart type, a cookbook of common chart patterns with styling defaults (colorblind palette, tight bounding box, 150 DPI), and guidance on interacting with non-coding users: when to ask for a data sample, how to handle iteration, when to flag a potentially misleading chart. The whole thing is under 400 lines so it fits in context whenever it triggers.
# the skill is just a Markdown file Claude reads on activation
git clone https://github.com/jkoets/Seaborn-Charts-Skill \
~/.claude/skills/seaborn-charts
pip install seaborn pandas openpyxl
Current Status
Working skill with a README and a USAGE walkthrough. Single-file and easy to tweak -; swap the default palette, change the theme, add work-specific recipes -; and Claude Code re-reads it each session, so edits take effect on the next conversation.
- Handles CSV, Excel, and DataFrame inputs.
- Decision table + chart cookbook with sane styling defaults baked in.
- Customizable: palette, theme, DPI, output format.