Визуализация данных и дашборды

★ 7.5 · design

dataviz is a Claude Code skill that provides a systematic, design-system-agnostic procedure for producing charts, dashboards, and any data visualization that is elegant, accessible, and consistent across light and dark modes. A validated default palette lives in `references/palette.md`, and a runnable `scripts/validate_palette.js` checks every categorical palette for CVD separation (ΔE ≥ 8 in OKLab), contrast, and lightness band — eliminating guesswork. The seven-step procedure covers form selection (bar, line, area, heatmap, scatter, sparkline, donut, stat tile, KPI row), color assignment by semantic role, mark specs, hover-layer interaction, and a final accessibility pass with legend rules and table-view fallback. Hard constraints — no dual-axis charts, fixed categorical hue order, sequential palettes in one hue only, no rainbow scales — are documented alongside an anti-patterns reference. Built for developers and designers who need analytically correct, brand-swappable data interfaces.