← The Wire

Public story · 2026-07-02 · high

Claude Code ships a dataviz skill with a color validator

The skill checks chart palettes for contrast and accessibility in both light and dark mode before code gets written.

Why now: Claude Code's changelog lists the /dataviz skill as of July 2, before the invoke-it-first habit has had time to form.

Confidence
high
Sources
1
Redaction
passed

Story

Claude Code added a /dataviz skill that checks chart color palettes for contrast before agents write visualization code, per the changelog.

Models get categorical colors and contrast wrong constantly on their own. A chart that fails contrast checks in dark mode still ships if nobody's watching for it.

The skill pairs chart-design guidance with a runnable palette validator. An agent can check a color set for contrast and accessibility in light and dark mode before generating chart code, instead of fixing colors afterward. The changelog says to invoke it before writing any chart code, not after.

That ordering is the whole point. A validator that catches a bad palette after the chart is coded means someone has to notice, flag it, and ask for a rewrite. Catching it before the first line of chart code exists skips that step entirely.

I've watched models pick categorical color sets that look right until you actually check contrast, especially against a dark background. That's a design-eye read, the kind you catch in a glance without needing a tool. A validator that catches it automatically, before the chart exists, is the kind of guardrail that should've shipped a long time ago.

This only works if people invoke it before writing chart code, not after a bad palette is already sitting in a dashboard or a PR. It's in Claude Code's changelog as of July 2, and that invoke-it-first habit hasn't formed yet for anyone.

Related stories

Each link below shares sources, entities, or timing with this story.

  1. Shared entities / Same source / Shared topic

    Claude Code Ships a /dataviz Skill for Charts

    Both cover Claude Code, Invoke; cite the same source (changelog); overlapping topics (catch, chart, claude, code, color).

  2. Shared entity: Claude Code / Same source / Shared topic / Earlier coverage

    Claude Code's June updates doubled rate limits and added nested skills.

    Both cover Claude Code; cite the same source (changelog); overlapping topics (agent, claude, code, skill).

  3. Claude Code v2.1.193 adds shell-classifier routing and response telemetry; the June 25 build kills two long-standing footguns.

    Both cover Claude Code; cite the same source (changelog); overlapping topics (agent, claude, code).

  4. Claude Code ships nested subagents, five levels deep.

    Both cover Claude Code; cite the same source (changelog); overlapping topics (agent, claude, code).

  5. Claude Code 2.1.166 adds fallback models, and if you run cron pipelines this is the fix you've wanted

    Both cover Claude Code; cite the same source (changelog); overlapping topics (agent, claude, code).

  6. Claude Code v2.1.194–195 adds trusted-device gating and a flag to kill mouse clicks.

    Both cover Claude Code; cite the same source (changelog); overlapping topics (claude, code).

  7. Claude Code 2.1.191 makes `/rewind` resume from before a `/clear`.

    Both cover Claude Code; cite the same source (changelog); overlapping topics (claude, code).

  8. Claude Code's fallbackModel chains up to three models for resilience.

    Both cover Claude Code; cite the same source (changelog); overlapping topics (claude, code).

Source trail

Entities

Claim evidence

  1. Claude Code ships a dataviz skill with a color validator

Provenance

Canonical issue
2026-07-02
AI generated
yes
Story unit
2026-07-02-claude-code-s-new-dataviz-skill-enforces-chart-and-palette-discipline
Labels
source-backed, canonical briefing excerpt