UI skill benchmark for AI coding agents

Same brief, same model.Only the skill changes.

SkillArena gives one model the same product briefs again and again, adding a different UI skill to the system prompt each time. Compare the pages side by side, starting with a no-skill control group.

01Skills
02Briefs
03Pages generated
04Model

/01Matrix

Every skill, every brief

One row per skill, one column per brief. The control group comes first. Open a skill to see all of its runs.

Loading the latest results…

/02About

A controlled experiment

01

Same brief

Every page answers the same product brief, word for word. Nothing in the task changes from one row to the next.

02

Same model, only the skill changes

All pages come from the same model. The only difference between rows is the skill added to the system prompt. The first row adds no skill and serves as the control group.

03

Three runs per pair

Each skill and brief pair runs three times, so you can tell a skill's real effect from a lucky generation.

Every link in the matrix opens the full results on skillarena.dev. Open skillarena.dev →

/03Waitlist

Get new results first

We'll email you when new skills, briefs or models land. No spam, and you can unsubscribe at any time.