本文へスキップ
Kitbay
スキル

programmatic-figure

Apache-2.02026年11月11日 更新261GitHubスタースクリプトあり

A skill from the anywhere-agents configuration project for drawing quantitative figures for papers and technical posts from data and runnable code, such as benchmark comparisons, ablations, scaling curves, distributions and heatmaps.

スキルの内容

A skill from the anywhere-agents configuration project for drawing quantitative figures for papers and technical posts from data and runnable code, such as benchmark comparisons, ablations, scaling curves, distributions and heatmaps. It uses Python and Matplotlib with a headless backend, or keeps a project's existing R or PGFPlots pipeline. A table maps the reader's question to a chart type, and the workflow checks the data for units and missing runs, keeps one order and one style per method, builds at the manuscript's column width and has the agent read the exported figure back to confirm values can be recovered. Output is a vector PDF or SVG plus a PNG preview, shipped with the builder script. Flowcharts and diagrams are left to a sibling skill.

できること

  • Pick a chart type for the question asked
  • Build publication figures with Matplotlib
  • Export vector PDF or SVG with a PNG preview
  • Check a rendered figure by reading it back