Scientific Figure Pro

Use this skill to create or revise publication figures with consistent aesthetics from the figures4papers project.

Workflow

1. Identify the figure family (bar, trend, heatmap, scatter, multi-panel, conceptual illustration).
2. Load helper APIs from scripts/scientific_figure_pro.py.
3. Apply style first with apply_publication_style(...).
4. Build the figure using make_* helpers.
5. Export with finalize_figure(...) to keep layout and output defaults consistent.

Use The Helper Module

Load by file path when package imports are inconvenient:

import importlib.util
from pathlib import Path
import matplotlib.pyplot as plt

module_path = Path("skills/scientific-figure-pro/scripts/scientific_figure_pro.py")
spec = importlib.util.spec_from_file_location("scientific_figure_pro", module_path)
mod = importlib.util.module_from_spec(spec)
spec.loader.exec_module(mod)

Core APIs

• FigureStyle(...): Configure typography and axis linewidth.
• apply_publication_style(style=None): Apply global rcParams.
• create_subplots(nrows, ncols, figsize, **kwargs): Create flattened axes.
• make_grouped_bar(...): Build grouped bars with publication defaults.
• make_trend(...): Build trend lines with optional confidence shadow.
• make_heatmap(...): Build heatmaps with optional labels and annotations.
• make_scatter(...): Build clean scatter plots.
• make_sphere_illustration(...): Build conceptual shaded sphere panels.
• finalize_figure(fig, out_path, formats=None, dpi=300, pad=0.05): Save outputs consistently.

Style Policy

• Use Helvetica/Arial-like sans-serif families.
• Remove top/right spines.
• Keep legends frameless.
• Use semantic palette mapping:
  • blue for target/proposed method,
  • green for improvements,
  • red for contrasts/baselines,
  • neutral for support categories.
• Use dpi=300 by default and dpi=600 for dense bar panels.

Practical Defaults

• Dense benchmark bars: FigureStyle(font_size=24, axes_linewidth=3) and wide canvas.
• Compact analysis charts: FigureStyle(font_size=15 or 16, axes_linewidth=2).
• Finalize every figure through finalize_figure(...) instead of direct plt.savefig(...).

Dependencies

• Python 3.8+
• matplotlib
• numpy

References

• Read references/design_theory.md when you need style rationale and reproduction rules.