Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Use when reviewing stylesheets, component styles, and responsive behavior related to Remove unused CSS rules. Check the rendered layout across breakpoints and interaction states before proposing a fix.
.claude/skills/thedaviddias-unused-css/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-07 | ✗→✓ | ▲ Improved | -71% | 0% |
| case-19 | ✓→✗ | ▼ Worse | 1% | 0% |
| case-01 | ✓→✓ | ＝ Same ✓ | 9% | 0% |
| case-02 | ✓→✓ | ＝ Same ✓ | -16% | 0% |
| case-03 | ✓→✓ | ＝ Same ✓ | 93% | 0% |
Most projects use only 10-25% of their CSS—removing the rest can cut bundle sizes by 75%+ and dramatically improve load times.
Use tools like PurgeCSS, UnCSS, or Chrome DevTools Coverage to identify and remove unused CSS to reduce bundle size and improve performance.
Implement automated unused CSS removal in your build process and regularly audit CSS files to eliminate dead code and unused selectors.
Explain how unused CSS bloats bundle sizes, slows page loading, and how modern tools can safely remove unused styles without breaking functionality.
Review stylesheets, component styles, and responsive states related to Remove unused CSS rules. Flag exact selectors, declarations, or breakpoints that violate the rule in the rendered UI.
For full implementation details, code examples, and framework-specific guidance, see references/rule.md.
Rule page: https://frontendchecklist.io/en/rules/css/unused-css
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | pass→pass | 8,650 | 7,662 | -11% | 1 | 1 | 0% | 1,567 | 1,704 | +9% | 0 | 0 | — |
case-02 | pass→pass | 10,190 | 7,249 | -29% | 1 | 1 | 0% | 1,920 | 1,606 | -16% | 0 | 0 | — |
case-03 | pass→pass | 3,246 | 4,211 | +30% | 1 | 1 | 0% | 518 | 1,000 | +93% | 0 | 0 | — |
case-04 | pass→pass | 7,114 | 7,971 | +12% | 1 | 1 | 0% | 1,324 | 1,693 | +28% | 0 | 0 | — |
case-05 | pass→pass | 15,940 | 11,477 | -28% | 1 | 1 | 0% | 2,559 | 2,167 | -15% | 0 | 0 | — |
case-06 | pass→pass | 10,765 | 6,356 | -41% | 1 | 1 | 0% | 1,702 | 1,439 | -15% | 0 | 0 | — |
case-07 | fail→pass | 11,191 | 2,542 | -77% | 1 | 1 | 0% | 1,951 | 568 | -71% | 0 | 0 | — |
case-08 | fail→fail | 13,602 | 8,396 | -38% | 1 | 1 | 0% | 2,305 | 1,586 | -31% | 0 | 0 | — |
case-09 | pass→pass | 4,920 | 4,846 | -2% | 1 | 1 | 0% | 876 | 1,083 | +24% | 0 | 0 | — |
case-10 | pass→pass | 11,058 | 10,232 | -7% | 1 | 1 | 0% | 2,020 | 2,313 | +15% | 0 | 0 | — |
case-11 | pass→pass | 13,287 | 11,818 | -11% | 1 | 1 | 0% | 2,061 | 2,136 | +4% | 0 | 0 | — |
case-12 | pass→pass | 11,311 | 8,060 | -29% | 1 | 1 | 0% | 1,735 | 1,565 | -10% | 0 | 0 | — |
case-13 | pass→pass | 12,160 | 7,044 | -42% | 1 | 1 | 0% | 2,319 | 1,713 | -26% | 0 | 0 | — |
case-14 | pass→pass | 9,230 | 6,409 | -31% | 1 | 1 | 0% | 1,726 | 1,423 | -18% | 0 | 0 | — |
case-15 | pass→pass | 16,035 | 13,972 | -13% | 1 | 1 | 0% | 2,530 | 2,530 | 0% | 0 | 0 | — |
case-16 | pass→pass | 12,996 | 10,916 | -16% | 1 | 1 | 0% | 2,277 | 2,017 | -11% | 0 | 0 | — |
case-17 | pass→pass | 10,339 | 7,421 | -28% | 1 | 1 | 0% | 1,711 | 1,510 | -12% | 0 | 0 | — |
case-18 | pass→pass | 8,879 | 6,812 | -23% | 1 | 1 | 0% | 1,539 | 1,417 | -8% | 0 | 0 | — |
case-19 | pass→fail | 15,456 | 13,386 | -13% | 1 | 1 | 0% | 2,489 | 2,504 | +1% | 0 | 0 | — |
case-20 | pass→pass | 8,441 | 6,686 | -21% | 1 | 1 | 0% | 1,445 | 1,472 | +2% | 0 | 0 | — |
case-21 | pass→pass | 8,607 | 5,952 | -31% | 1 | 1 | 0% | 1,433 | 1,359 | -5% | 0 | 0 | — |
case-22 | pass→pass | 14,008 | 12,171 | -13% | 1 | 1 | 0% | 2,440 | 2,362 | -3% | 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 0 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.