Imported from projectious-work/aibox (context/templates/processkit/v0.23.0/context/skills/documents/infographics/SKILL.md). Install upstream with npx skills add projectious-work/aibox --skill infographics. Copyright stays with the author.

Infographics

Intro

A good chart starts from a single message, picks the most accurate visual encoding for the data relationship, and strips everything that does not support it. Generate standalone SVG with explicit viewBox, semantic text, and accessible metadata so the result is editable, responsive, and screen-reader friendly.

Overview

SVG generation principles

• Emit standalone SVG with xmlns="http://www.w3.org/2000/svg".
• Set an explicit viewBox (e.g. viewBox="0 0 800 600") for responsive scaling.
• Use width="100%" and height="auto" for flexible embedding.
• Group related elements with <g> and position via transform.
• Put reusable styles in <defs> and <style> blocks.
• Give elements semantic IDs and classes so the file stays editable.
• Keep text as <text> elements, not paths, for accessibility.

Data-to-visual mapping

1. Identify the message — what should the viewer learn?
2. Choose the encoding — position, length, angle, area, color, or shape.
3. Apply visual hierarchy — the most important data gets the strongest visual treatment.
4. Add context — title, axis labels, legend, source attribution.
5. Simplify — remove anything that does not support the message.

Encoding effectiveness from most to least accurate for quantitative data: position > length > angle > area > color saturation > shape.

Chart type selection

Visual design

• Color palette: 3-5 colors max, legible in grayscale.
• Typography: one sans-serif font, clear size hierarchy (title > subtitle > labels > annotations).
• Whitespace: generous margins and padding prevent clutter.
• Grid lines: light gray, thin, removed when unnecessary.
• Data labels: place directly on chart elements when space permits to reduce legend lookups.
• Aspect ratio: ~16:9 for presentations, ~4:3 for documents, ~1:1 for social media.
• Mobile first: over 60% of content is consumed on mobile — prefer vertical layouts.
• One focal point: visuals with a clear focal point outperform complex ones by roughly 2x in retention.

Accessibility

• 4.5:1 contrast for text, 3:1 for large text and UI elements.
• Never rely solely on color to convey information — add patterns, labels, or shapes.
• Include <title> and <desc> for screen readers.
• Use role="img" and aria-label on the SVG element.

Alternative formats

When SVG is not the right fit, consider Mermaid for flowcharts and sequence/Gantt diagrams embeddable in Markdown, ASCII art for terminals and plain-text READMEs, or a CSV plus narrative when the user just needs organized data with commentary.

Gotchas

Agent-specific failure modes — provider-neutral pause-and-self-check items:

• Omitting the viewBox attribute. An SVG without viewBox does not scale responsively — it renders at a fixed pixel size that overflows narrow containers and documentation embeds. Always emit viewBox="0 0 W H" with explicit dimensions and use width="100%" and height="auto".
• Positioning <text> elements without explicit coordinates and alignment attributes. SVG text without x, y, text-anchor, and dominant-baseline renders at the origin or at browser-specific defaults. Lay out every text element explicitly with all four positioning attributes.
• Encoding data in color alone. Roughly 8% of men have red-green color vision deficiency. A chart that differentiates categories only by color is inaccessible. Always pair color with a second encoding: pattern, label, shape, or position.
• Choosing a pie chart for more than 5 slices. The human eye is poor at comparing angles; with many slices, most segments become indistinguishable. Prefer a horizontal bar chart, which exposes differences via length — the most accurate quantitative encoding.
• Omitting <title> and <desc> elements. Screen readers have no way to announce the chart's content without these elements. Include a <title> stating the chart's finding (not just its topic) and a <desc> summarizing the key data.
• Starting a bar chart's y-axis at a non-zero value. Truncating the axis to exaggerate small differences distorts the viewer's perception of magnitude. For bar charts, the y-axis baseline must be zero.
• Hardcoding width and height pixel attributes on the root SVG element. A chart with fixed pixel dimensions overflows its container on mobile and cannot be embedded responsively. Use width="100%" and height="auto" and let viewBox drive the aspect ratio.

Full reference

Tufte's core rules

1. Show the data — maximize the data-ink ratio (data-ink / total ink).
2. Avoid chartjunk — decorative elements that do not convey information.
3. Use small multiples — repeat a chart design across subsets for comparison.
4. Integrate text and graphics — labels belong on the chart, not in a distant legend.

Cleveland & McGill encoding ranking

Perceptual accuracy, most to least accurate for quantitative data:

1. Position along a common scale
2. Position along non-aligned scales
3. Length
4. Angle / slope
5. Area
6. Volume / curvature
7. Color saturation / shading
8. Color hue

Always pick the highest-ranked encoding that fits the relationship.

Color palettes

• Sequential (low to high): single hue, varying lightness (e.g. #f7fbff to #08306b). Good for heatmaps, choropleths, and continuous data.
• Diverging (negative to positive with neutral center): two hues with light center (red-white-blue, brown-white-teal). Good for deviation from a baseline.
• Categorical: maximally distinct hues at similar lightness. Limit to 5-7 colors. A safe set: #4e79a7 #f28e2b #e15759 #76b7b2 #59a14f #edc948 #b07aa1.
• Colorblind safety (~8% of men): avoid red-green as the only cue. Test with deuteranopia and protanopia simulators. Safe pairs: blue/orange, blue/red, purple/green.

Typography scale

SVG layout template

<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 800 500"
     role="img" aria-label="Chart description">
  <title>Chart Title</title>
  <desc>Detailed description for screen readers</desc>

  <style>
    .title { font: bold 24px sans-serif; fill: #1a1a1a; }
    .subtitle { font: 16px sans-serif; fill: #666; }
    .axis-label { font: 14px sans-serif; fill: #333; }
    .tick-label { font: 12px sans-serif; fill: #666; }
    .grid-line { stroke: #e0e0e0; stroke-width: 1; }
    .data-label { font: 12px sans-serif; fill: #333; }
    .source { font: 10px sans-serif; fill: #999; }
  </style>

  <!-- Margin convention: top=50, right=30, bottom=60, left=80 -->
  <g transform="translate(80, 50)">
    <!-- Chart content here -->
    <!-- Plot area: 690 x 390 -->
  </g>

  <text x="80" y="30" class="title">Chart Title</text>
  <text x="80" y="48" class="subtitle">Supporting context</text>
  <text x="80" y="490" class="source">Source: Data source</text>
</svg>

Consistent margins for an 800x500 canvas: top 50 (title and subtitle), right 30 (breathing room), bottom 60 (x-axis label plus tick labels), left 80 (y-axis label plus tick labels).

Chart-specific guidelines

• Bar charts: gap between bars 40-50% of bar width. Sort by value descending unless order is meaningful. Horizontal bars for long category labels. Always start the y-axis at zero.
• Line charts: cap at 5-7 lines before it gets cluttered. Label lines directly when possible. Add dots at data points for fewer than 15 points. Consider area fills for 1-2 series.
• Scatter plots: use 0.3-0.7 opacity when points overlap. Size by area, not radius. Add a trend line only when the correlation is meaningful. Label outliers directly.
• Tables (when better than charts): use when precise values matter more than patterns. Right-align numbers, left-align text. Subtle row striping for readability. Highlight key values with bold or color.

Common pitfalls

• Truncated y-axis: non-zero start exaggerates differences.
• 3D effects: distort value perception — always 2D.
• Pie charts: hard to compare — bar charts are more precise.
• Dual y-axes: confusing — use small multiples instead.
• Excessive decoration: chartjunk distracts from data.
• Poor contrast: light colors on white backgrounds.
• Missing units: always label what the numbers mean.
• Rainbow colormaps: perceptually non-uniform — use sequential or diverging palettes.

Responsive considerations

Use viewBox without fixed width and height for fluid scaling. Test readability at 50% and 200% zoom. Keep minimum effective text size around 10px. On mobile, stack small multiples vertically and prefer horizontal bars.

References

• references/best-practices.md — Tufte, Cleveland & McGill, color palettes, SVG template, and chart-specific guidelines source.