Claude Code Can Now Audit Your Figma Design System. Here's Why That's a Big Deal.
Claude Code's /design plugin can do more than generate screens. It can read your existing Figma design system, flag inconsistencies, and suggest extensions. For designers building their own apps, that changes the audit game entirely.
Claude Code's /design plugin has quietly grown a new trick: it can audit an existing Figma design system and suggest how to extend it, not just generate fresh screens from scratch. If you've ever spent a Friday afternoon hunting for rogue colour values or duplicate components, you already know why that matters.
The shift from "generate" to "understand"
Most AI design tools have been pointed at blank canvases. You describe a screen, the tool draws one. Useful, but limited, because real product work is mostly about working within an existing system, not inventing one from nothing.
The workflow described by Nick Babich on UX Planet signals something different. The /design plugin, used inside Claude Code, can now be directed at a design system you already own. The use case is a design system audit followed by refinement, meaning Claude reads what's there and tells you what needs fixing or filling in. That's a fundamentally different job, and it's one designers have historically had to do by hand or with clunky spreadsheet-based audits.
What a design system audit with Claude Code actually looks like
The source article sits behind a Medium paywall, so the full step-by-step isn't publicly available. Based on what is visible, here's what the workflow involves at a high level.
You trigger the audit inside Claude Code using the /design plugin. Claude then examines the design system, looking for gaps, inconsistencies, or components that could be standardised. From there, the conversation moves into refinement, which the author has elsewhere described as a three-stage loop: design critique, UX writing review, and design handoff prep.
Think of it like hiring a very thorough junior designer to go through your Figma library and produce a written audit report, except it happens in minutes and you can ask follow-up questions in plain language.
For a designer building their own app without a dedicated design-systems team, this is meaningful. Audits usually get skipped because they're tedious. If Claude can shoulder most of that tedium, you're more likely to actually keep your system tidy as your product grows.
How to try this yourself right now
A few practical pointers for getting started, grounded in what the article and its companion pieces describe.
Start with the /design plugin inside Claude Code. This is not the same as talking to Claude in the browser chat. Claude Code is a separate, more powerful environment aimed at building real things. If you haven't set it up yet, that's the first step.
Point it at a specific scope, not your whole library at once. Design systems can be enormous. Ask Claude to audit a single component family first, say your button variants or your type scale. A narrow scope gives you a legible output you can actually act on.
Use the critique-to-handoff loop. The companion workflow the author describes goes: design critique (what's broken or inconsistent), UX writing review (are labels and microcopy consistent), then handoff prep (is everything named and documented clearly enough for a developer to use). Running those three stages in order gives you a clean, ship-ready component by the end.
Treat the output as a starting point, not a verdict. Claude will flag things that look inconsistent by pattern. Whether they're actually a problem depends on context you hold, not the AI. Review every suggestion before acting on it.
What this doesn't solve
A couple of honest caveats are worth naming.
The full article is paywalled, so the precise prompts and exact Claude Code commands used aren't confirmed here. Treat the workflow above as directional rather than a guaranteed recipe.
Claude Code is also still a technical environment. It's more approachable than writing code by hand, but it isn't a no-setup tool. Expect a learning curve if you're coming to it cold.
And design system quality still depends on the decisions you make. Claude can surface that two shades of grey look nearly identical, but it can't tell you which one carries the brand intent. That judgment stays with you.
The grounded takeaway
Design system audits have always been the unglamorous backbone of good product work. They're important, they're skipped constantly, and they compound into real problems when a product scales. The idea that Claude Code's /design plugin can now assist with this job, not just with greenfield screen generation, represents a genuine shift in how a solo designer or small team can maintain quality without a dedicated systems role. Worth experimenting with, carefully and with your own judgment firmly in the loop.