Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Use to write or review microcopy — CTAs, error messages, empty states, confirmation dialogs, tooltips, onboarding text.
.claude/skills/getcrew44-ux-copy/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-04 | ✗→✓ | ▲ Improved | 39% | 0% |
| case-12 | ✗→✓ | ▲ Improved | 106% | 0% |
| case-13 | ✗→✓ | ▲ Improved | -27% | 0% |
| case-20 | ✗→✓ | ▲ Improved | 211% | 0% |
| case-23 | ✓→✗ | ▼ Worse | 87% | 0% |
Words in interfaces are design decisions. They set tone, frame action, and decide how the user feels at the moment they read them.
Structure: what happened + why + how to fix.
"Payment declined. Your card was declined by your bank. Try a different card or contact your bank." Avoid blaming the user. Avoid "An error occurred." Avoid stack-trace nouns ("ValidationError: field invalid").
Structure: what this is + why it's empty + how to start.
"No projects yet. Create your first project to start collaborating with your team." Empty states are onboarding moments — they're the first impression for users who haven't done anything yet.
Concise, helpful, never restating what's already obvious from the label. If the tooltip says the same thing as the label, cut the tooltip.
Set expectations. "Loading…" is fine for under a second. "Processing your file — this can take 30 seconds" is fine when it actually takes 30 seconds. Lying about duration erodes trust.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-05 | pass→pass | 7,733 | 8,221 | +6% | 1 | 1 | 0% | 1,206 | 1,970 | +63% | 0 | 0 | — |
case-14 | pass→pass | 19,926 | 17,202 | -14% | 1 | 1 | 0% | 3,446 | 3,597 | +4% | 0 | 0 | — |
case-15 | pass→pass | 10,297 | 9,869 | -4% | 1 | 1 | 0% | 1,681 | 2,223 | +32% | 0 | 0 | — |
case-01 | pass→pass | 7,899 | 7,720 | -2% | 1 | 1 | 0% | 1,244 | 1,997 | +61% | 0 | 0 | — |
case-02 | pass→pass | 8,416 | 7,105 | -16% | 1 | 1 | 0% | 1,280 | 1,806 | +41% | 0 | 0 | — |
case-03 | pass→pass | 8,443 | 7,687 | -9% | 1 | 1 | 0% | 1,389 | 2,001 | +44% | 0 | 0 | — |
case-04 | fail→pass | 8,882 | 8,242 | -7% | 1 | 1 | 0% | 1,429 | 1,983 | +39% | 0 | 0 | — |
case-06 | pass→pass | 4,610 | 7,149 | +55% | 1 | 1 | 0% | 664 | 1,863 | +181% | 0 | 0 | — |
case-07 | pass→pass | 6,423 | 8,278 | +29% | 1 | 1 | 0% | 899 | 1,989 | +121% | 0 | 0 | — |
case-08 | pass→pass | 11,020 | 8,087 | -27% | 1 | 1 | 0% | 1,720 | 1,920 | +12% | 0 | 0 | — |
case-09 | pass→pass | 7,834 | 9,902 | +26% | 1 | 1 | 0% | 1,215 | 2,218 | +83% | 0 | 0 | — |
case-10 | pass→pass | 5,610 | 5,753 | +3% | 1 | 1 | 0% | 885 | 1,543 | +74% | 0 | 0 | — |
case-11 | pass→pass | 6,981 | 8,300 | +19% | 1 | 1 | 0% | 1,055 | 1,970 | +87% | 0 | 0 | — |
case-12 | fail→pass | 6,705 | 8,914 | +33% | 1 | 1 | 0% | 1,054 | 2,173 | +106% | 0 | 0 | — |
case-13 | fail→pass | 20,012 | 10,043 | -50% | 1 | 1 | 0% | 3,118 | 2,291 | -27% | 0 | 0 | — |
case-16 | pass→pass | 16,886 | 13,725 | -19% | 1 | 1 | 0% | 2,385 | 2,777 | +16% | 0 | 0 | — |
case-17 | pass→pass | 5,976 | 6,929 | +16% | 1 | 1 | 0% | 1,066 | 1,793 | +68% | 0 | 0 | — |
case-18 | pass→pass | 9,843 | 9,124 | -7% | 1 | 1 | 0% | 1,434 | 2,132 | +49% | 0 | 0 | — |
case-19 | pass→pass | 10,271 | 9,757 | -5% | 1 | 1 | 0% | 1,581 | 2,247 | +42% | 0 | 0 | — |
case-20 | fail→pass | 4,344 | 9,366 | +116% | 1 | 1 | 0% | 694 | 2,157 | +211% | 0 | 0 | — |
case-21 | pass→pass | 5,627 | 7,131 | +27% | 1 | 1 | 0% | 831 | 1,849 | +123% | 0 | 0 | — |
case-22 | pass→pass | 5,256 | 9,875 | +88% | 1 | 1 | 0% | 919 | 2,185 | +138% | 0 | 0 | — |
case-23 | pass→fail | 9,021 | 14,278 | +58% | 1 | 1 | 0% | 1,697 | 3,167 | +87% | 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. The headline lift of +13 percentage points is the difference between those two pass rates over the 23 comparable cases. 1 case got worse with the skill loaded, and it is 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.