UI skills for AI coding agents, compared
Same brief. Same model. Only the skill changes.
Every page here was generated by … from the same product briefs. Each row adds one UI skill to the system prompt; the first row adds none, as a control group. Compare them side by side.
- 01Briefs…fixed
- 02Model…fixed
- 03Skill…variable
- 04Output…screenshot
The dataset
One controlled experiment, rendered as a grid.
Source: skillarena.dev live data · loading…
Matrix
Every skill × every brief.
Rows are skills, starting with the no-skill control. Columns are briefs. Open a skill to see its full results on skillarena.dev.
| Skill |
|---|
← Scroll sideways to see every brief →
No skill matches “”.
About
A fair comparison means holding everything else still.
Same brief, same model
Every page is built from the same product brief by the same model, …. Nothing about the prompt changes between rows except the skill.
Only the skill changes
Each row adds one skill to the system prompt. The first row adds no skill at all, so you can see what the model does on its own.
Three runs per pair
Each skill and brief pair is generated three times, so one lucky or unlucky run does not decide the result. The thumbnails link to the full results on skillarena.dev.
What a thumbnail is, and is not. A thumbnail is a screenshot of a generated page. It shows output, not quality scores. A blank cell means no screenshot is available for that pair.
Waitlist
Get updates when new skills and briefs land.
Join the list to hear when the matrix grows. Tell us your role so we know who is using the comparisons.
- Who receives itThe SkillArena team, through the waitlist endpoint on skillarena.dev.
- What we storeYour email address, your role, and where you came from.
- What happens nextYou get SkillArena updates by email. Nothing else.