Our review
Scaffolds matplotlib figure code for publication-ready TDA visuals following the conventions of a specific codebase.
Strengths
- Enforces shared constants (sizes, colors, rcParams) for visual consistency.
- Automatically saves both PDF and PNG at the correct resolution.
- Provides specialized templates for persistence diagrams, barcodes, null histograms, and landscapes.
- Clearly separates production figures from exploratory ones.
Limitations
- Requires the `trajectory_tda` package to be installed and set up.
- Only generates skeleton code, not a fully rendered figure.
- Strict annotation rules may not fit all contexts.
When you need to produce a standardized figure for a paper or presentation using the `trajectory_tda` codebase.
If you are working outside this ecosystem or need highly customized figure styles not covered by the templates.
Security analysis
SafeThe skill contains only code-generation instructions for matplotlib figures following a style guide. It does not execute any commands, access the network, or perform potentially harmful actions.
No concerns found
Examples
/tda-figure-spec persistence-diagram P01-B/tda-figure-spec null-histogram P01-A/tda-figure-spec/tda-figure-spec — Generate Publication-Ready TDA Figures
Scaffold matplotlib figure code following the established trajectory_tda/viz/ conventions.
Uses PUBLICATION_RC, DPI, FIGSIZE_*, STATE_COLORS, and _save_figure from the
codebase — never ad-hoc sizes or colours.
Usage
/tda-figure-spec [figure-type] [paper-number]
Example: /tda-figure-spec persistence-diagram P01-B
Example: /tda-figure-spec null-histogram P01-A
Example: /tda-figure-spec (interactive)
Style constants (always import from trajectory_tda/viz/constants.py)
from trajectory_tda.viz.constants import (
DPI, FIGSIZE_FULL, FIGSIZE_WIDE, FIGSIZE_HALF, FIGSIZE_SQUARE,
PUBLICATION_RC, STATE_COLORS, STATES, STATE_LABELS, REGIME_LABELS,
)
import matplotlib.pyplot as plt
plt.rcParams.update(PUBLICATION_RC)
Figure size guide
| Use case | Constant | Size |
|---|---|---|
| Single full-width plot | FIGSIZE_FULL | 190mm × 120mm |
| Wide two-panel | FIGSIZE_WIDE | 190mm × 104mm |
| Half-width single panel | FIGSIZE_HALF | 90mm × 120mm |
| Square heatmap | FIGSIZE_SQUARE | 90mm × 90mm |
Save function (always use this pattern)
def _save_figure(fig, output_dir: Path, name: str) -> None:
output_dir.mkdir(parents=True, exist_ok=True)
fig.savefig(output_dir / f"{name}.pdf", format="pdf")
fig.savefig(output_dir / f"{name}.png", format="png", dpi=DPI)
Colour rules
- State space: always
STATE_COLORSdict — nevertab10or default cycle - Observations:
#2d6a2e(green); Null:steelblue; Dim 0: green; Dim 1:#1a237e - Regimes:
plt.cm.tab10withREGIME_LABELSorder - Sequential/diverging:
viridis,plasma,RdBu_r— neverjet
Annotation standards
- p-values: write
p=0.003, notp<0.01 - Axes: remove top and right spines; keep bottom and left
- Legends:
frameon=False - Subplot labels:
ax.text(0.02, 0.97, "(a)", transform=ax.transAxes, ...)
Figure templates available
| Template | When to use |
|---|---|
| plot_persistence_diagram(diagram, dim, ax) | Scatter birth-death pairs |
| plot_barcode(diagram, dim, ax) | Feature lifetimes sorted by persistence |
| plot_null_histogram(observed, null_dist, p_value, dim, ax) | Permutation test result |
| plot_landscape_comparison(obs, null, grid, ax) | L² landscape distance (mandatory) |
Output paths
| Context | Path | Naming |
|---|---|---|
| Production paper figure | papers/PXX/figures/ | fig{N}_{desc}.{pdf\|png} |
| Supplement | papers/PXX/figures/ | figS{N}_{desc}.{pdf\|png} |
| Working / exploratory | figures/{domain}/ | YYYYMMDD_{desc}.png |
Working figures must not be saved to papers/PXX/figures/ — that is for production only.
Prompt Engineering
Data & AI
Prompt engineering best practices and templates to maximize AI outputs.
Data Visualization
Data & AI
Generates data visualizations and charts tailored to your data.
RAG Architecture Setup
Data & AI
Setup guide for RAG (Retrieval-Augmented Generation) architectures.