Agent Auto Sci Data Viz
Use this subskill whenever data, tables, or figures are central to the output.
Fast Workflow
- Inspect file types and schema.
- Audit missingness, outliers, duplicates, ranges, and units.
- Define each table or figure by the claim it should support.
- Choose statistical tests and effect sizes before writing results.
- Use publication-safe palettes and legible dimensions.
- Export reproducible figures and record code paths.
- 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-analysisstatistical-analysismatplotlibseabornscientific-visualizationnetworkxxlsx
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.