Scientific Figure Generation
Generate publication-quality figures for research papers.
Input
$0 — Description of the desired figure$1 — (Optional) Path to data file (CSV, JSON, NPY, PKL) or results directory
Scripts
Generate figure template
python ~/.claude/skills/figure-generation/scripts/figure_template.py --type bar --output figure_script.py --name comparison
python ~/.claude/skills/figure-generation/scripts/figure_template.py --list-types
Available types: bar, training-curve, heatmap, ablation, line, scatter, radar, violin, tsne, attention
Three-Phase Pipeline (from MatPlotAgent)
Phase 1: Query Expansion
Expand the user's figure description into step-by-step coding specifications using the prompts in references/figure-prompts.md. Determine: figure type, data mapping (x/y/color/hue), style requirements, paper conventions.
Phase 2: Code Generation with Execution Loop (up to 4 retries)
- Generate a self-contained Python script using the template from
scripts/figure_template.py as a starting point
- Write script to a temp file and execute:
python figure_script.py
- If error: capture traceback, feed back, regenerate (see ERROR_PROMPT in references)
- If no
.png produced: add explicit save instruction, retry
- On success: report the generated figure path
Phase 3: Visual Refinement
Read the generated PNG file and visually inspect using the VLM feedback prompts from references/figure-prompts.md:
- Does the figure type match the request?
- Are labels, titles, and legends correct?
- Is the color scheme appropriate and consistent?
- Are axis scales sensible? Is text readable at publication size?
If improvements needed: generate corrective instructions and re-execute.
References
- All MatPlotAgent prompts:
~/.claude/skills/figure-generation/references/figure-prompts.md
- Figure templates:
~/.claude/skills/figure-generation/scripts/figure_template.py
Output
Both PNG (preview, 300 DPI) and PDF (vector, for paper) formats. Plus the LaTeX include code:
\begin{figure}[t]
\centering
\includegraphics[width=\linewidth]{figures/figure_name.pdf}
\caption{Description. Best viewed in color.}
\label{fig:figure_name}
\end{figure}
Quality Requirements
- DPI ≥ 300, or vector PDF
- Colorblind-friendly palette (no red-green only)
- All text ≥ 8pt at print size
- Consistent styling across all paper figures
- No matplotlib default title — use LaTeX caption
Related Skills
1---2name: figure-generation3description: Generate publication-quality scientific figures using matplotlib/seaborn with a three-phase pipeline (query expansion, code generation with execution, VLM visual feedback). Handles bar charts, line plots, heatmaps, training curves, ablation plots, and more. Use when the user needs figures, plots, or visualizations for a paper.4---56# Scientific Figure Generation78Generate publication-quality figures for research papers.910## Input1112- `$0` — Description of the desired figure13- `$1` — (Optional) Path to data file (CSV, JSON, NPY, PKL) or results directory1415## Scripts1617### Generate figure template18```bash19python ~/.claude/skills/figure-generation/scripts/figure_template.py --type bar --output figure_script.py --name comparison20python ~/.claude/skills/figure-generation/scripts/figure_template.py --list-types21```2223Available types: `bar`, `training-curve`, `heatmap`, `ablation`, `line`, `scatter`, `radar`, `violin`, `tsne`, `attention`2425## Three-Phase Pipeline (from MatPlotAgent)2627### Phase 1: Query Expansion28Expand the user's figure description into step-by-step coding specifications using the prompts in `references/figure-prompts.md`. Determine: figure type, data mapping (x/y/color/hue), style requirements, paper conventions.2930### Phase 2: Code Generation with Execution Loop (up to 4 retries)311. Generate a self-contained Python script using the template from `scripts/figure_template.py` as a starting point322. Write script to a temp file and execute: `python figure_script.py`333. If error: capture traceback, feed back, regenerate (see ERROR_PROMPT in references)344. If no `.png` produced: add explicit save instruction, retry355. On success: report the generated figure path3637### Phase 3: Visual Refinement38Read the generated PNG file and visually inspect using the VLM feedback prompts from `references/figure-prompts.md`:39- Does the figure type match the request?40- Are labels, titles, and legends correct?41- Is the color scheme appropriate and consistent?42- Are axis scales sensible? Is text readable at publication size?4344If improvements needed: generate corrective instructions and re-execute.4546## References4748- All MatPlotAgent prompts: `~/.claude/skills/figure-generation/references/figure-prompts.md`49- Figure templates: `~/.claude/skills/figure-generation/scripts/figure_template.py`5051## Output5253Both PNG (preview, 300 DPI) and PDF (vector, for paper) formats. Plus the LaTeX include code:5455```latex56\begin{figure}[t]57 \centering58 \includegraphics[width=\linewidth]{figures/figure_name.pdf}59 \caption{Description. Best viewed in color.}60 \label{fig:figure_name}61\end{figure}62```6364## Quality Requirements65- DPI ≥ 300, or vector PDF66- Colorblind-friendly palette (no red-green only)67- All text ≥ 8pt at print size68- Consistent styling across all paper figures69- No matplotlib default title — use LaTeX caption7071## Related Skills72- Upstream: [data-analysis](../data-analysis/), [experiment-code](../experiment-code/)73- Downstream: [paper-writing-section](../paper-writing-section/), [paper-compilation](../paper-compilation/), [slide-generation](../slide-generation/)74- See also: [table-generation](../table-generation/)
om-scogo