Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Audits the intersection of web performance and accessibility. Covers lazy loading, skeleton screens, CLS impact on AT, code splitting loading states, and progressive enhancement patterns.
.claude/skills/community-access-performance-accessibility/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-22 | ✗→✓ | ▲ Improved | 5% | 0% |
| case-01 | ✗→✓ | ▲ Improved | -8% | 0% |
| case-02 | ✗→✓ | ▲ Improved | -3% | 0% |
| case-04 | ✓→✗ | ▼ Worse | -19% | 0% |
| case-07 | ✓→✓ | ＝ Same ✓ | 4% | 0% |
Derived from .claude/agents/performance-accessibility.md. Treat platform-specific tool names or delegation instructions as Codex equivalents.
You audit where web performance optimization intersects with accessibility. Performance techniques can introduce accessibility barriers if not implemented carefully.
alt, size placeholders (prevent CLS), announce content arrival, "Load more" button for infinite scrollaria-hidden="true" on skeletons, aria-busy="true" on container, announce load completionaspect-ratioprefers-reduced-motion, CSS over JS animations, disableable parallaxfont-display: swap, above-fold accessible immediately| Performance | Risk | Solution | |-------------|------|----------| | Lazy images | Missing alt on placeholders | Preserve alt, size placeholder | | Infinite scroll | Keyboard trap | "Load more" button alternative | | Skeleton screens | SR reads placeholders | aria-hidden + aria-busy | | Code splitting | Flash of inaccessible content | Accessible loading state |
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-07 | pass→pass | 14,377 | 14,405 | +0% | 1 | 1 | 0% | 2,427 | 2,533 | +4% | 0 | 0 | — |
case-21 | pass→pass | 14,128 | 9,309 | -34% | 1 | 1 | 0% | 2,399 | 1,882 | -22% | 0 | 0 | — |
case-22 | fail→pass | 17,204 | 16,476 | -4% | 1 | 1 | 0% | 2,647 | 2,773 | +5% | 0 | 0 | — |
case-08 | pass→pass | 14,200 | 9,178 | -35% | 1 | 1 | 0% | 2,207 | 1,661 | -25% | 0 | 0 | — |
case-06 | pass→pass | 14,826 | 14,284 | -4% | 1 | 1 | 0% | 2,330 | 2,533 | +9% | 0 | 0 | — |
case-01 | fail→pass | 21,541 | 17,669 | -18% | 1 | 1 | 0% | 3,559 | 3,276 | -8% | 0 | 0 | — |
case-02 | fail→pass | 22,717 | 19,818 | -13% | 1 | 1 | 0% | 3,806 | 3,681 | -3% | 0 | 0 | — |
case-03 | pass→pass | 11,066 | 9,191 | -17% | 1 | 1 | 0% | 1,973 | 1,836 | -7% | 0 | 0 | — |
case-04 | pass→fail | 15,126 | 9,374 | -38% | 1 | 1 | 0% | 2,237 | 1,821 | -19% | 0 | 0 | — |
case-05 | pass→pass | 13,133 | 14,656 | +12% | 1 | 1 | 0% | 2,245 | 2,612 | +16% | 0 | 0 | — |
case-09 | pass→pass | 13,937 | 13,471 | -3% | 1 | 1 | 0% | 2,263 | 2,434 | +8% | 0 | 0 | — |
case-10 | pass→pass | 16,433 | 15,990 | -3% | 1 | 1 | 0% | 2,665 | 2,834 | +6% | 0 | 0 | — |
case-11 | pass→pass | 13,530 | 11,762 | -13% | 1 | 1 | 0% | 2,213 | 2,319 | +5% | 0 | 0 | — |
case-12 | pass→pass | 12,697 | 11,006 | -13% | 1 | 1 | 0% | 2,250 | 2,242 | -0% | 0 | 0 | — |
case-13 | pass→pass | 12,118 | 9,248 | -24% | 1 | 1 | 0% | 1,891 | 1,836 | -3% | 0 | 0 | — |
case-14 | pass→pass | 4,890 | 3,664 | -25% | 1 | 1 | 0% | 739 | 881 | +19% | 0 | 0 | — |
case-15 | pass→pass | 16,118 | 13,793 | -14% | 1 | 1 | 0% | 2,656 | 2,610 | -2% | 0 | 0 | — |
case-16 | pass→pass | 17,121 | 15,436 | -10% | 1 | 1 | 0% | 2,524 | 2,537 | +1% | 0 | 0 | — |
case-17 | pass→pass | 6,383 | 3,933 | -38% | 1 | 1 | 0% | 1,045 | 843 | -19% | 0 | 0 | — |
case-18 | pass→pass | 14,308 | 9,894 | -31% | 1 | 1 | 0% | 2,493 | 2,137 | -14% | 0 | 0 | — |
case-19 | pass→pass | 9,573 | 7,520 | -21% | 1 | 1 | 0% | 1,398 | 1,479 | +6% | 0 | 0 | — |
case-20 | pass→pass | 12,003 | 11,950 | -0% | 1 | 1 | 0% | 2,082 | 2,527 | +21% | 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 +9 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.