Agent Auto Sci Data Viz

Use this subskill whenever data, tables, or figures are central to the output.

Fast Workflow

1. Inspect file types and schema.
2. Audit missingness, outliers, duplicates, ranges, and units.
3. Define each table or figure by the claim it should support.
4. Choose statistical tests and effect sizes before writing results.
5. Use publication-safe palettes and legible dimensions.
6. Export reproducible figures and record code paths.
7. Write captions that stand alone.

Read references/data_analysis_visualization_workflows.md.

For full paper projects, this skill owns EDA, statistical-test reporting, robustness visualization, table shells, publication figure planning, figure-title/caption checks, and final visual evidence audits. It should work with agent-auto-sci-geospatial for maps and with agent-auto-sci-scicomm for figure narrative.

For deeper K-Dense-style encapsulation:

• references/k_dense_data_viz_mapping.md: how EDA, statistics, scientific visualization, matplotlib, seaborn, and network analysis are adapted.
• references/review_bibliometric_figure_system.md: figure system for review, bibliometric, policy agenda, and sport geography empirical manuscripts.

Related Helper Skills

• exploratory-data-analysis
• statistical-analysis
• matplotlib
• seaborn
• scientific-visualization
• networkx
• xlsx

Must Not Do

• Do not make decorative charts that do not answer a research question.
• Do not choose tests after seeing desired significance.
• Do not use color palettes that fail grayscale or colorblind checks.
• Do not let map/figure aesthetics override evidence clarity.