SkillArena
Controlled comparison

Same brief. Same model. Only the skill changes.

SkillArena compares UI skills for AI coding agents. Every page here was generated by one model from one product brief; each row adds a different skill to the system prompt, and the first row adds none at all.

  1. 1One briefEvery skill gets the identical product brief.
  2. 2One modelThe same model writes every page.
  3. 3One skill per rowThe no-skill control group sets the baseline.
—
Skills
—
Briefs
—
Pages
—
Model

The matrix

One row per skill, one column per brief. The no-skill control group comes first. Select a skill to see its full results.

Loading the matrix…

About the experiment

Skills promise better interfaces. SkillArena tests that promise with a controlled setup, so differences you see come from the skill and nothing else.

i.

Same brief, same model

Each brief is given word-for-word to the same model. The only variable is the skill added to the system prompt.

ii.

Three runs per pair

Each skill and brief pair is generated three times, so one lucky or unlucky run doesn't decide the result.

iii.

Full results online

Every skill row links to its full results on skillarena.dev, with all runs and pages side by side.

Join the waitlist

Get an update when new skills, briefs and models are added to the arena. No spam, only the results.