Does your design system reach your AI tools?

Give an agent three ordinary briefs, once without your system and once with it, then count what it got wrong. This page walks you through it, one step at a time.

3 briefs
2 runs
~30 min
any agent

Here's what you'll do

Same three briefs, same agent, twice: once in an empty project, once with your design system. The difference is what your system is worth to the agent. What's left wrong in the second run is what to fix next.

02Set up: stack and retired list
03Run A: without the system
04Run B: with your system
05Compare: a vs b, side by side
06Score Run B: count the mistakes
07Your result: what to fix first

Check you have these

I've picked the agent and model my team actually usesClaude Code, Cursor, v0, Lovable, Figma Make… whatever you'd use on a normal day.
I can open our real repo or workspace with the system set upRules files, MCP servers, skills or a connected library, the way the team has them now.
I have about 30 minutes and room for two windows side by sideRun A and Run B can run in parallel.

3 left on this step. You can come back to them.