Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Write the small UI text that guides users — buttons, labels, tooltips, CTAs, confirmations. Use when asked to write microcopy, button/CTA text, form labels, tooltips, helper text, or to make UI wording clearer. Produces specific, action-oriented microcopy with options and rationale, matched to the moment and the product's voice — concise, scannable, and free of jargon.
.claude/skills/mohitagw15856-microcopy-writer/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-02 | ✗→✓ | ▲ Improved | 25% | 0% |
| case-03 | ✗→✓ | ▲ Improved | 9% | 0% |
| case-06 | ✗→✓ | ▲ Improved | 27% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 29% | 0% |
| case-11 | ✗→✓ | ▲ Improved | 93% | 0% |
Microcopy is the smallest text with the biggest leverage: a button label, a field hint, a confirmation. Good microcopy is clear, action-oriented, and reduces hesitation — it tells the user exactly what will happen and what to do. This skill writes that text for a specific moment, with a couple of options and the reasoning, so the team can choose with intent.
Given "a button for the checkout step" or a screenshot description, write the microcopy anyway — infer the context, the user's goal, and the voice, and label assumptions. Offer 2–3 options where wording is a judgement call. Never hand back a question instead of copy.
Ask for these only if they aren't already provided (else infer and label):
For each piece of text:
Apply the principles: lead with a verb for actions ("Save changes", not "OK"); say what happens next; keep it short and specific; match voice; and for risky/irreversible actions, make the consequence explicit ("Delete 3 files" beats "Are you sure?"). Cover the related states if relevant (default, loading, success, error).
End with consistency notes — terms/patterns to reuse elsewhere so the product speaks with one voice.
UX writing practice — action-oriented, expectation-setting microcopy, voice consistency, and clarity at decision points.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 12,851 | 9,103 | -29% | 1 | 1 | 0% | 2,004 | 2,508 | +25% | 0 | 0 | — |
case-02 | fail→pass | 9,411 | 7,911 | -16% | 1 | 1 | 0% | 1,692 | 2,116 | +25% | 0 | 0 | — |
case-03 | fail→pass | 10,777 | 8,598 | -20% | 1 | 1 | 0% | 2,031 | 2,212 | +9% | 0 | 0 | — |
case-04 | pass→pass | 25,083 | 7,128 | -72% | 1 | 1 | 0% | 1,135 | 1,996 | +76% | 0 | 0 | — |
case-05 | pass→pass | 8,280 | 20,649 | +149% | 1 | 1 | 0% | 1,438 | 1,883 | +31% | 0 | 0 | — |
case-06 | fail→pass | 10,914 | 8,583 | -21% | 1 | 1 | 0% | 1,630 | 2,073 | +27% | 0 | 0 | — |
case-07 | fail→pass | 10,642 | 24,752 | +133% | 1 | 1 | 0% | 1,737 | 2,247 | +29% | 0 | 0 | — |
case-08 | pass→pass | 19,179 | 26,140 | +36% | 1 | 1 | 0% | 1,033 | 1,941 | +88% | 0 | 0 | — |
case-09 | pass→pass | 7,314 | 18,182 | +149% | 1 | 1 | 0% | 1,386 | 2,004 | +45% | 0 | 0 | — |
case-10 | pass→pass | 5,243 | 8,410 | +60% | 1 | 1 | 0% | 935 | 2,195 | +135% | 0 | 0 | — |
case-11 | fail→pass | 6,581 | 8,742 | +33% | 1 | 1 | 0% | 1,285 | 2,484 | +93% | 0 | 0 | — |
case-12 | pass→pass | 6,245 | 7,313 | +17% | 1 | 1 | 0% | 1,197 | 2,093 | +75% | 0 | 0 | — |
case-13 | fail→pass | 7,173 | 9,157 | +28% | 1 | 1 | 0% | 1,385 | 2,408 | +74% | 0 | 0 | — |
case-14 | pass→pass | 3,131 | 6,603 | +111% | 1 | 1 | 0% | 505 | 1,755 | +248% | 0 | 0 | — |
case-15 | pass→pass | 6,172 | 8,186 | +33% | 1 | 1 | 0% | 1,168 | 2,166 | +85% | 0 | 0 | — |
case-16 | fail→pass | 8,464 | 8,755 | +3% | 1 | 1 | 0% | 1,579 | 2,259 | +43% | 0 | 0 | — |
case-17 | pass→pass | 8,142 | 8,691 | +7% | 1 | 1 | 0% | 1,546 | 2,275 | +47% | 0 | 0 | — |
case-18 | fail→pass | 7,894 | 8,816 | +12% | 1 | 1 | 0% | 1,365 | 2,146 | +57% | 0 | 0 | — |
case-19 | pass→pass | 8,060 | 8,601 | +7% | 1 | 1 | 0% | 1,462 | 2,219 | +52% | 0 | 0 | — |
case-20 | pass→fail | 28,128 | 14,024 | -50% | 1 | 1 | 0% | 5,250 | 3,215 | -39% | 0 | 0 | — |
case-21 | pass→pass | 12,324 | 13,110 | +6% | 1 | 1 | 0% | 2,082 | 2,976 | +43% | 0 | 0 | — |
case-22 | pass→pass | 10,081 | 6,253 | -38% | 1 | 1 | 0% | 2,260 | 2,039 | -10% | 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. 22 cases were attempted. The headline lift of +32 percentage points is the difference between those two pass rates over the 22 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.