| name | python-plotting-skill |
| description | Use when Codex needs to choose, generate, review, or export Python scientific figures with Matplotlib from described data, tables, arrays, CSV-like structures, or synthetic examples; use for chart selection, runnable plotting scripts, PNG/SVG export, and figure limitation notes. |
Python Plotting Skill
Use this skill for Python scientific figures when the user wants a static,
reproducible plot rather than an interactive dashboard.
Workflow
- Identify the communication task: trend, comparison, relationship,
distribution, matrix, uncertainty, or repeated panels.
- Inspect the data shape. If the data is private, ask for a small synthetic
sample with the same columns and rough ranges.
- Choose a template from
scripts/render_gallery.py.
- Generate a runnable Matplotlib script with fixed seeds for examples.
- Export PNG and SVG unless the user asks otherwise.
- Explain what the chart can show and what it cannot prove.
- Before public use, run
bash scripts/release_check.sh in the repository or
an equivalent privacy/provenance check.
Template map
line_trend: one ordered series.
multi_line_comparison: several series on the same x axis.
scatter_relationship: two numeric variables.
regression_scatter: x-y relation with a simple fitted line.
confidence_band: mean and uncertainty band.
grouped_bar: grouped categorical comparison.
heatmap_matrix: matrix or grid values.
density_scatter: dense scatter with point density coloring.
box_jitter: distribution plus observations.
violin_plot: distribution shape by group.
small_multiples: repeated panels.
category_small_multiples: repeated category panels.
correlation_matrix: pairwise correlations.
lollipop_ranking: sorted item rankings.
paired_before_after: paired changes between two conditions.
spectral_density: frequency content for sampled signals.
residual_convergence: residual decay across solver or simulation iterations.
main_inset: main trend with one magnified local region.
shared_colorbar_panels: several heatmaps compared on one shared color scale.
gantt_timeline: simple task timing on a calendar axis.
Boundaries
Do not copy private data, paper figures, local paths, unclear snippets, or
third-party gallery assets into generated examples. Use synthetic data when the
goal is to demonstrate a visual pattern.
For diagrams, use scientific-diagram-skill. For MATLAB render workflows, use
matlab-plotting-skill.