The shadcn/ui project released lint, an MIT-licensed linter for Tailwind design systems, according to the project's repository. It lets teams write rules that an AI coding agent, not just a human reviewer, can verify and fix against.
The tool defines six rules, including no-restyle, no-raw-colors and no-arbitrary-values, and lets teams write component-specific contracts that control which styling properties each component accepts, such as permitting a Button to take margin classes but not padding overrides, per the README. It works with Tailwind v4 projects with or without shadcn/ui and runs on both ESLint and Oxlint.
When a rule is violated, the error message explains what went wrong and suggests fixes drawn from the design system's actual components, variants and theme values. The project's own testing found that coding agents given that error feedback resolved most violations after a single correction round.
Design systems have always assumed a human eventually reads the linter output. Tooling built so an agent can read the same error, understand the constraint and fix its own code is a small sign of what a codebase has to look like once agents are doing a meaningful share of the styling work.