Blog-post figures

Figures for the web are read in a scrolling column, on phones, without a caption nearby. They must carry their own headline, explain their own encoding, and survive being screenshotted. This skill encodes the design language of a set of strong research-blog figures (catalogued in references/gallery.md) and ships code that produces it.

Two routes, chosen by content:

Content	Route	Tooling
Data charts: bars, lines, step/survival curves, lollipops, scatter, multi-panel grids	matplotlib	`scripts/blogfig.py` helpers
Process/flow diagrams, timelines, quote/transcript cards, heatmap tables, dashboards of cards	HTML + CSS	`assets/figure.css` + templates, `scripts/render_html.py`

The skill root is the directory containing this file (e.g. ~/.claude/skills/blogpost-figure; find ~ -name blogfig.py -path '*blogpost-figure*' 2>/dev/null if unsure). Call scripts by absolute path. Python scripts are uv run scripts with inline deps; for your own figure script use uv run --with matplotlib --with numpy python my_fig.py or a PEP 723 header. For LaTeX/PDF paper figures use the paper-figure skill instead.