Your Design System Has a New User, and It Can’t See Your Figma File
Nishant Choudhary · 25 October 2026
Read the originalNishant Choudhary, a design engineer, argues that a design system needs text rules and failing checks in the code, because AI coding agents cannot see files in Figma.
Nishant Choudhary is a design engineer and frontend architect who writes about design systems on Medium. He argues that AI coding agents now build more interface than most teams review, and that they read only the text they find in a code repository. He sets out five changes that make a design system usable by an AI agent, rather than only by people who can see the design tool Figma.
- Design tokens, the named values for colour and spacing, should be named for their purpose and listed in one short rules file that the agent always reads.
- Each component needs a written contract: its allowed variants as a closed list, and the uses it forbids, such as two primary buttons side by side.
- Rules that matter should fail the build, for example a lint rule against raw colour values, so that the agent reads the error and fixes its own code.
- He keeps a few reference screens in the repository and requires every data view to handle loading, empty, error and overflowing text.
A design-system team can move the rules a reviewer used to enforce by eye into files and checks that an AI agent reads. This is an opinion essay with code examples, without measured results.
Write a short contract for each design-system component in its code, listing the allowed variants and the uses it forbids, so that AI coding agents follow it when building screens.
Add a lint rule to the front-end build that fails when component code uses a raw hex colour instead of a design-system colour token, so that AI coding agents correct it.
Derived by Working Surface from the article; more in the Playbook. Source line: Documentation is a request. A failing check is a rule.
29 October 2026: as AI agents raise the volume of work to check, teams move the check closer to the work: fixed rules for what agents may fix, design rules written where agents read them, reviewers trained on how AI-written code goes wrong, and leaders who do the work and are reviewed in public.