Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Edit or add component theme classes and styles for base and carbon variants. Use when changing component appearance, Tailwind classes, theme tokens, Divider/Button themes, or files matching *.theme.ts.
.claude/skills/indielayer-component-theme/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-06 | ✗→✓ | ▲ Improved | -51% | 0% |
| case-01 | ✗→✓ | ▲ Improved | -41% | 0% |
| case-04 | ✗→✓ | ▲ Improved | -47% | 0% |
| case-05 | ✗→✓ | ▲ Improved | -62% | 0% |
| case-07 | ✗→✓ | ▲ Improved | -70% | 0% |
For component <Name>:
theme/<Name>.base.theme.tstheme/<Name>.carbon.theme.tsImport *Theme from ../<Name>.vue.
tsconst theme: ButtonTheme = { classes: { wrapper({ props, slots, data }) { return ['...'] }, }, styles({ props, colors, css, data }) { /* CSS variables */ }, } export default theme
Export in both:
packages/ui/src/themes/base/components.tspackages/ui/src/themes/carbon/components.tsKey must match the theme name used in useTheme('<Name>', ...).
When changing layout, states, or tokens in one variant, apply the same logical change to the other unless the carbon design intentionally differs.
Reference: packages/ui/src/components/button/theme/Button.base.theme.ts
defaultsVariant/layout booleans that should respect app.use(UI, { defaults }) must use optionalBooleanProp() from packages/ui/src/common/props.ts (not plain Boolean). Use useResolvedComponentProps when the template or script reads those props outside useTheme.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-06 | fail→pass | 17,238 | 5,913 | -66% | 1 | 1 | 0% | 2,714 | 1,319 | -51% | 0 | 0 | — |
case-01 | fail→pass | 30,133 | 16,340 | -46% | 1 | 1 | 0% | 4,963 | 2,921 | -41% | 0 | 0 | — |
case-02 | fail→fail | 18,476 | 10,329 | -44% | 1 | 1 | 0% | 3,310 | 2,378 | -28% | 0 | 0 | — |
case-03 | fail→fail | 4,009 | 10,871 | +171% | 1 | 1 | 0% | 486 | 1,733 | +257% | 0 | 0 | — |
case-04 | fail→pass | 13,091 | 3,845 | -71% | 1 | 1 | 0% | 1,991 | 1,062 | -47% | 0 | 0 | — |
case-05 | fail→pass | 13,836 | 3,253 | -76% | 1 | 1 | 0% | 1,997 | 754 | -62% | 0 | 0 | — |
case-07 | fail→pass | 23,847 | 4,978 | -79% | 1 | 1 | 0% | 3,149 | 945 | -70% | 0 | 0 | — |
case-08 | fail→pass | 11,302 | 4,682 | -59% | 1 | 1 | 0% | 2,012 | 1,000 | -50% | 0 | 0 | — |
case-09 | fail→pass | 18,875 | 2,544 | -87% | 1 | 1 | 0% | 2,619 | 616 | -76% | 0 | 0 | — |
case-10 | pass→pass | 4,728 | 2,483 | -47% | 1 | 1 | 0% | 909 | 667 | -27% | 0 | 0 | — |
case-20 | pass→pass | 15,198 | 11,012 | -28% | 1 | 1 | 0% | 2,664 | 2,508 | -6% | 0 | 0 | — |
case-11 | fail→pass | 21,073 | 4,304 | -80% | 1 | 1 | 0% | 2,909 | 981 | -66% | 0 | 0 | — |
case-12 | fail→pass | 26,056 | 4,697 | -82% | 1 | 1 | 0% | 2,962 | 1,069 | -64% | 0 | 0 | — |
case-13 | fail→pass | 9,733 | 2,151 | -78% | 1 | 1 | 0% | 1,377 | 503 | -63% | 0 | 0 | — |
case-14 | fail→pass | 13,733 | 2,917 | -79% | 1 | 1 | 0% | 2,003 | 790 | -61% | 0 | 0 | — |
case-15 | fail→fail | 13,715 | 2,888 | -79% | 1 | 1 | 0% | 2,060 | 769 | -63% | 0 | 0 | — |
case-16 | fail→pass | 16,191 | 6,307 | -61% | 1 | 1 | 0% | 3,523 | 1,543 | -56% | 0 | 0 | — |
case-17 | pass→pass | 12,567 | 4,373 | -65% | 1 | 1 | 0% | 2,136 | 1,126 | -47% | 0 | 0 | — |
case-18 | fail→pass | 9,556 | 3,815 | -60% | 1 | 1 | 0% | 1,672 | 894 | -47% | 0 | 0 | — |
case-19 | fail→fail | 45,901 | 1,680 | -96% | 1 | 1 | 0% | 8,109 | 520 | -94% | 0 | 0 | — |
case-21 | pass→pass | 16,362 | 12,034 | -26% | 1 | 1 | 0% | 3,466 | 2,159 | -38% | 0 | 0 | — |
case-22 | pass→pass | 23,577 | 11,553 | -51% | 1 | 1 | 0% | 1,771 | 1,782 | +1% | 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 +59 percentage points is the difference between those two pass rates over the 22 comparable cases.
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.