Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Review UI code for Web Interface Guidelines compliance by the Vercel engineering team. Covers layout, typography, color, motion, and accessibility for product UI.
.claude/skills/nexu-io-web-design-guidelines/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-13 | ✗→✓ | ▲ Improved | -24% | 0% |
| case-21 | ✗→✓ | ▲ Improved | -28% | 0% |
| case-05 | ✓→✗ | ▼ Worse | -73% | 0% |
| case-06 | ✓→✗ | ▼ Worse | -80% | 0% |
| case-04 | ✓→✓ | ＝ Same ✓ | 43% | 0% |
Review files for compliance with Web Interface Guidelines.
references/guidelines.md — this is the default, reproducible execution pathfile:line formatDefault (pinned): The vendored copy at references/guidelines.md is the default execution path. It is a pinned snapshot of the upstream guidelines and guarantees the skill produces reproducible results in any runtime environment.
Upstream (live): For users who want the very latest rules, the live source is available at:
https://raw.githubusercontent.com/vercel-labs/web-interface-guidelines/main/command.mdFetching the live version is optional — the pinned snapshot is always the baseline.
When a user provides a file or pattern argument:
references/guidelines.md as the default sourceIf no files specified, ask the user which files to review.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 4,551 | 16,481 | +262% | 1 | 1 | 0% | 677 | 499 | -26% | 0 | 0 | — |
case-02 | fail→fail | 6,375 | 6,522 | +2% | 1 | 1 | 0% | 357 | 491 | +38% | 0 | 0 | — |
case-03 | fail→fail | 7,969 | 6,055 | -24% | 1 | 1 | 0% | 1,238 | 581 | -53% | 0 | 0 | — |
case-04 | pass→pass | 19,018 | 18,323 | -4% | 1 | 1 | 0% | 3,026 | 4,342 | +43% | 0 | 0 | — |
case-05 | pass→fail | 16,294 | 7,444 | -54% | 1 | 1 | 0% | 2,336 | 633 | -73% | 0 | 0 | — |
case-06 | pass→fail | 15,499 | 10,569 | -32% | 1 | 1 | 0% | 3,244 | 635 | -80% | 0 | 0 | — |
case-07 | fail→fail | 17,910 | 5,214 | -71% | 1 | 1 | 0% | 1,571 | 594 | -62% | 0 | 0 | — |
case-08 | fail→fail | 6,726 | 10,231 | +52% | 1 | 1 | 0% | 1,008 | 1,631 | +62% | 0 | 0 | — |
case-09 | pass→pass | 12,702 | 3,102 | -76% | 1 | 1 | 0% | 1,788 | 773 | -57% | 0 | 0 | — |
case-10 | fail→fail | 12,057 | 5,762 | -52% | 1 | 1 | 0% | 2,016 | 1,191 | -41% | 0 | 0 | — |
case-11 | fail→fail | 5,885 | 9,114 | +55% | 1 | 1 | 0% | 891 | 662 | -26% | 0 | 0 | — |
case-12 | fail→fail | 29,738 | 7,247 | -76% | 1 | 1 | 0% | 2,393 | 764 | -68% | 0 | 0 | — |
case-13 | fail→pass | 8,063 | 5,162 | -36% | 1 | 1 | 0% | 1,077 | 814 | -24% | 0 | 0 | — |
case-14 | fail→fail | 15,649 | 6,091 | -61% | 1 | 1 | 0% | 2,183 | 624 | -71% | 0 | 0 | — |
case-15 | fail→fail | 10,916 | 6,099 | -44% | 1 | 1 | 0% | 1,624 | 876 | -46% | 0 | 0 | — |
case-16 | fail→fail | 9,995 | 11,029 | +10% | 1 | 1 | 0% | 1,705 | 1,587 | -7% | 0 | 0 | — |
case-17 | fail→fail | 14,516 | 6,731 | -54% | 1 | 1 | 0% | 1,535 | 878 | -43% | 0 | 0 | — |
case-18 | fail→fail | 6,189 | 7,329 | +18% | 1 | 1 | 0% | 880 | 737 | -16% | 0 | 0 | — |
case-19 | pass→pass | 5,986 | 2,355 | -61% | 1 | 1 | 0% | 915 | 662 | -28% | 0 | 0 | — |
case-20 | fail→fail | 11,360 | 7,022 | -38% | 1 | 1 | 0% | 1,477 | 793 | -46% | 0 | 0 | — |
case-21 | fail→pass | 19,521 | 12,868 | -34% | 1 | 1 | 0% | 2,654 | 1,899 | -28% | 0 | 0 | — |
case-22 | fail→fail | 8,633 | 5,852 | -32% | 1 | 1 | 0% | 1,172 | 611 | -48% | 0 | 0 | — |
case-23 | fail→fail | 15,623 | 10,008 | -36% | 1 | 1 | 0% | 2,158 | 1,334 | -38% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 23 cases were attempted, and 13 counted toward the lift figure. The other 10 produced results that are not comparable between the two arms, so they are excluded from the headline rather than averaged into it. The headline lift of 0 percentage points is the difference between those two pass rates over the 13 comparable cases. 2 cases got worse with the skill loaded, and they are included in that figure.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.