agentleFS
AvinavaAvinava/document-design-system/skills/chart-design/SKILL.md

chart-design

Skill2 starsChanged 36 days agoMITView on GitHub

Design honest, accessible charts of quantitative data as inline SVG — bar and column charts, line and time series, limit ledgers, distributions, scatter plots, small multiples, and waterfall bridges. Use when visualizing measured values, choosing a chart type, picking or fixing chart colors, building a categorical palette from design tokens, setting axes and scales, labelling series, or making an existing chart readable in grayscale and in print. Do not use for diagrams of structure or process such as architecture or flows (use diagram-design), for the surrounding report narrative and metric semantics (use analytical-document-design), or for interactive dashboards that require a client-side charting library.

2 other projects use a version of this.

Adopted by

Most recently taken up by

  • Other versions2 projects use a version of this.

    Word for word · 0

    None found.

    Edited · 0

    None found.

    Wrote their own version · 2

    1. adityareddy-dev/deskstudio★ 43same name
    2. thewyattbrocato/design-and-copy-skills★ 2same name

    How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.

  • In this repository6 other files it gives its agents.

The file

We split this skill into the rules it gives an agent

12 rules. About 0 are common, 11 are distinctive, and 1 is very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈11

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 1

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Equal visual weight means equal comparability: Do not put two different units on one axis.

    line 20
  2. Distinctive:

    If the data does not support the chart's implicit claim, change the chart, not the axis.

    line 25
  3. Distinctive:

    Never a pie chart with more than three slices.

    line 42
  4. Distinctive:

    Never a dual-axis chart. The crossover point is set by the axis scaling, which means it is set by the author, which means it proves nothing.

    line 44
  5. Distinctive:

    Categorical palettes are a last resort: If the chart needs more than one hue, it usually needs to be small multiples instead.

    line 55
  6. Distinctive:

    Verify in grayscale: If two series become indistinguishable, they were only ever distinguished by hue.

    line 58
  7. Distinctive:

    Show the value: when there are fewer than about twelve bars.

    line 67
  8. Very specific:

    Use the shared SVG classes in core/base.css (.bar-focal, .bar-comparison, .axis-line, .grid-line, .chart-label, .chart-value) rather than …

    line 90
  9. Distinctive:

    No more than eight visible categories.

    line 96
  10. Distinctive:

    No more than about five lines on one chart.

    line 97
  11. Distinctive:

    No chartjunk: no 3D, no shadows, no gradients, no textured fills, no background images.

    line 99
  12. Distinctive:

    Remove the axis line where the gridlines already imply it.

    line 100

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Chart Design

A chart is an argument about numbers. Its job is to make one comparison faster than a table would — and to make the wrong reading hard.

Choose a chart only when it reveals a relationship faster than a table. A chart of four values is slower to read than four labelled numbers, because the reader has to decode the encoding before they get the data.

The honesty rules

These are not stylistic. Breaking them changes what the reader concludes.

  • Bar and column charts start at zero. A truncated bar axis multiplies apparent differences by an arbitrary factor. If small differences matter, use a line chart, a dot plot, or state the values.
  • Line charts may use a non-zero axis — they encode position, not length — but the axis must be labelled and the truncation visible.
  • Area encodes one quantity, not two. Doubling both width and height quadruples the area for a doubled value.
  • Percentages name their denominator, near the chart, not buried in a footnote.
  • Equal visual weight means equal comparability. Do not put two different units on one axis.
  • Time on an axis is continuous. Missing months appear as zero or as a marked gap, never by silently closing up.
  • State when a scale is logarithmic, in the axis label, not the caption.
  • Categories are categories, even when they look like numbers. Years, quarters, ports, and version numbers are labels, not measurements. Put them on a categorical (band) scale. Left on a continuous scale, 2020…2025 gets positioned by arithmetic distance, and a missing year silently becomes a gap the data never had.

If the data does not support the chart's implicit claim, change the chart, not the axis.

Choosing a form

The reader needsForm
Rank and compare categoriesHorizontal bars, sorted descending
Compare a value across time periodsColumns, chronological
A trend, or many series over timeLine
Used versus available against a limitLimit ledger (single track)
Composition of a whole, 2–3 partsStacked bar, labelled
Composition of a whole, many partsRanked bars — not a pie
How a total is built from its partsWaterfall bridge — parts must sum to the total
Relationship between two measuresScatter
The same comparison across groupsSmall multiples
Distribution shapeHistogram or box plot

Never a pie chart with more than three slices. Angle is the hardest encoding to judge, and a ranked bar chart answers the same question in less space.

Never a dual-axis chart. The crossover point is set by the axis scaling, which means it is set by the author, which means it proves nothing.

Per-form construction rules are in references/chart-forms.md.

Color

The token contract in core/tokens.md applies. Charts add one constraint that matters more here than anywhere else in this system:

Color is the last encoding to reach for, and never the only one.

  • Default to one accent plus muted ink. One focal bar in var(--accent), everything else in var(--comparison-fill). This handles most analytical charts and is the strongest hierarchy available.
  • Categorical palettes are a last resort. If the chart needs more than one hue, it usually needs to be small multiples instead.
  • Never encode a category by hue alone. Direct-label the series, or vary line style and marker shape as well.
  • Status colors are for status. var(--positive), var(--warning), var(--critical) mean something. Using them as a three-category palette destroys that meaning everywhere else in the document.
  • Verify in grayscale. If two series become indistinguishable, they were only ever distinguished by hue.

For deriving a genuine categorical palette from the theme when one is unavoidable, see references/palette.md.

Labels

Most chart failures are labelling failures.

  • Direct-label series at the end of a line or beside a bar. A legend is a lookup table the reader holds in memory.
  • Show the value when there are fewer than about twelve bars. The chart shows the shape; the number shows the fact.
  • Never put labels inside bars. They collide with short bars and get clipped.
  • Axis labels carry units. Units (thousands), not Units.
  • The title states the finding, not the variables. "Ingestion holds 41% of the footprint" beats "Size by function".
  • Name the as-of date and the denominator in or beside the figure.

Output

Inline SVG. No client-side charting library, unless interactivity is an explicit requirement — and it usually is not, because a static chart is what survives being printed, emailed, and screenshotted.

<figure class="chart">
  <svg role="img" aria-labelledby="ch-fn-title ch-fn-desc" viewBox="0 0 720 320" width="100%">
    <title id="ch-fn-title">Current footprint by function</title>
    <desc id="ch-fn-desc">Nine functions ranked by size. Ingestion holds 41 percent of the 1.8 million unit total, more than the next three combined.</desc>
    ...
  </svg>
  <figcaption>Source: normalized inventory export, as of 2026-08-12. Denominator: 1,842,110 included units.</figcaption>
</figure>

<title> first, unique IDs prefixed per figure, viewBox always. The <desc> states the comparison and its result — that sentence is what a screen-reader user gets instead of the chart, and writing it is the fastest test of whether the chart has a point.

Use the shared SVG classes in core/base.css (.bar-focal, .bar-comparison, .axis-line, .grid-line, .chart-label, .chart-value) rather than presentation attributes, so a theme change reaches every chart.

To generate charts from data rather than hand-authoring them, scripts/render_chart.mjs renders Observable Plot to a static SVG at authoring time. See references/rendering.md.

Density

  • No more than eight visible categories. Combine the tail into a labelled "Other" and keep every row in the supporting data.
  • No more than about five lines on one chart. Beyond that, use small multiples.
  • Gridlines are hairlines in var(--rule), or absent. They orient; they are not part of the data.
  • No chartjunk: no 3D, no shadows, no gradients, no textured fills, no background images.
  • Remove the axis line where the gridlines already imply it.

Before delivering

  • The chart is faster to read than the equivalent table.
  • Bars start at zero; any truncated axis is labelled and visible.
  • Every percentage names its denominator.
  • Series are direct-labelled, not legend-only.
  • No meaning is carried by hue or lightness alone; legends name the encoding in words.
  • Legible in grayscale.
  • The title states the finding.
  • viewBox, <title> first, <desc>, unique IDs.
  • Eight categories or fewer visible; the tail is accounted for.
  • Readable at print width, not just at authoring width.

Paths in this skill

core/… and scripts/… are relative to this skill's directory — the folder that contains this SKILL.md — not to your working directory. Each skill carries its own copy, so it works however it was installed. Prefix them with that directory:

node "<skill-dir>/scripts/render_chart.mjs" spec.json --out chart.svg

Reference files

  • references/chart-forms.md — per-form construction rules and failure modes.
  • references/palette.md — deriving a categorical palette from theme tokens, and when not to.
  • references/rendering.md — generating charts from data with Observable Plot.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.