Data Visualization Expert

The user's request: "$ARGUMENTS"

Critical Constraints

All shared rules in ../../assets/core-standards.md apply — accessibility (no color-only encoding, WCAG AA contrast, CVD-safe palettes), integrity (zero-baseline bars, no 3D, no dual y-axes, avoid pies), the 10% highlighting rule, and action titles. Read that file before designing.

Hebrew/Arabic text: only the text is RTL — chart geometry stays left-to-right (numbers are LTR in Hebrew too). Follow the RTL section in core-standards and the RTL recipe in references/code-recipes.md; never mirror axes or bar order.

Instructions

Step 1: Determine the Medium

Table, graph, or simple text — use the decision logic at the top of references/chart-types.md. If a table is the right medium, switch to the table-design skill.

Step 2: Choose the Chart Type

Match the data relationship to the visual form using the selection matrix in references/chart-types.md.

Step 3: Maximize the Data-Ink Ratio

Strip non-data elements (borders, background fills, dense gridlines); push axes and reference lines to light grey. Apply the rules in references/design-principles.md.

Step 4: Apply Branding and Color

Follow the branding workflow in ../../assets/core-standards.md: brand assets if present, otherwise grey baseline + one CVD-safe highlight.

Step 5: Focus Attention

Action title, axis units and time grain, annotate the key pattern directly on the chart, direct labels instead of legends (≤5 series).

Step 6: Accessibility Check

Verify deterministically — do not eyeball contrast:

python3 "${CLAUDE_PLUGIN_ROOT}/scripts/check_contrast.py" '#FG' '#BG'
python3 "${CLAUDE_PLUGIN_ROOT}/scripts/check_palette.py" '#hex1,#hex2,...'
python3 "${CLAUDE_PLUGIN_ROOT}/scripts/check_rtl.py" 'the chart title'   # if any text is Hebrew/Arabic

Add redundant encoding where color carries meaning; draft alt text for web/BI output. For RTL text, render a test image and read one label back — reversed glyph order is invisible in code.

Step 7: Validate

Run the "where are your eyes drawn?" test (see core-standards). If the gaze misses the insight, return to Step 3.

Step 8: Generate Output

Produce the chart in the user's preferred library, starting from the matching theme in references/code-recipes.md (matplotlib, plotly, vega-lite) so defaults comply by construction. After generating, run the visualization-review skill on the output (P0 checks at minimum).

Troubleshooting

Spaghetti graph (too many overlapping lines): highlight one key series in a vivid color and push the rest to grey, or split into small multiples sharing the same axes.