Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Audit a UI or design against WCAG 2.2 AA/AAA and ARIA patterns, returning criterion-referenced findings with severity and specific fixes. Use when the user wants an accessibility check, contrast verification, keyboard/screen-reader review, or wants to confirm a component meets POUR.
.claude/skills/plugin87-a11y-audit/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-13 | ✗→✓ | ▲ Improved | 128% | 0% |
| case-18 | ✗→✓ | ▲ Improved | -48% | 0% |
| case-19 | ✗→✓ | ▲ Improved | 95% | 0% |
| case-20 | ✓→✗ | ▼ Worse | -66% | 0% |
| case-21 | ✓→✗ | ▼ Worse | 73% | 0% |
Evaluate against WCAG 2.2 and the project's ARIA patterns.
accessibility/wcag-checklist.md (POUR-organized, P0/P1/P2) and accessibility/aria-patterns.md.node scripts/measure_render.mjs <file> [--dark] (every text element) AND node scripts/verify_states.mjs <file> [--dark] (every interactive element in default/hover/focus — catches hover-state failures). For loose color pairs, python3 scripts/contrast.py "<fg>" "<bg>". Never state a ratio you did not measure.taste/motion-choreography.md).A findings table: WCAG criterion (e.g. 1.4.3) · severity (P0/P1/P2) · what fails · specific fix. Confirm passes explicitly. Accessibility may never be traded for aesthetics.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-02 | fail→fail | 19,060 | 8,712 | -54% | 1 | 1 | 0% | 3,746 | 575 | -85% | 0 | 0 | — |
case-01 | fail→fail | 20,359 | 31,183 | +53% | 1 | 1 | 0% | 3,735 | 5,550 | +49% | 0 | 0 | — |
case-03 | fail→fail | 31,732 | 5,184 | -84% | 1 | 1 | 0% | 6,191 | 573 | -91% | 0 | 0 | — |
case-04 | fail→fail | 9,181 | 15,705 | +71% | 1 | 1 | 0% | 1,919 | 2,786 | +45% | 0 | 0 | — |
case-05 | fail→fail | 11,448 | 7,420 | -35% | 1 | 1 | 0% | 1,718 | 644 | -63% | 0 | 0 | — |
case-06 | fail→fail | 10,352 | 7,516 | -27% | 1 | 1 | 0% | 1,642 | 941 | -43% | 0 | 0 | — |
case-07 | pass→pass | 17,301 | 26,366 | +52% | 1 | 1 | 0% | 3,199 | 4,981 | +56% | 0 | 0 | — |
case-08 | fail→fail | 10,767 | 4,301 | -60% | 1 | 1 | 0% | 1,541 | 528 | -66% | 0 | 0 | — |
case-09 | fail→fail | 12,119 | 9,712 | -20% | 1 | 1 | 0% | 1,725 | 538 | -69% | 0 | 0 | — |
case-10 | fail→fail | 13,953 | 5,800 | -58% | 1 | 1 | 0% | 2,546 | 566 | -78% | 0 | 0 | — |
case-11 | fail→fail | 7,171 | 19,613 | +174% | 1 | 1 | 0% | 1,180 | 3,662 | +210% | 0 | 0 | — |
case-12 | fail→fail | 14,837 | 25,611 | +73% | 1 | 1 | 0% | 2,502 | 4,473 | +79% | 0 | 0 | — |
case-13 | fail→pass | 12,720 | 25,475 | +100% | 1 | 1 | 0% | 2,321 | 5,300 | +128% | 0 | 0 | — |
case-14 | fail→fail | 8,974 | 5,597 | -38% | 1 | 1 | 0% | 1,402 | 536 | -62% | 0 | 0 | — |
case-15 | fail→fail | 10,603 | 12,337 | +16% | 1 | 1 | 0% | 1,813 | 1,873 | +3% | 0 | 0 | — |
case-16 | fail→fail | 10,012 | 5,190 | -48% | 1 | 1 | 0% | 1,500 | 589 | -61% | 0 | 0 | — |
case-17 | fail→fail | 12,065 | 5,847 | -52% | 1 | 1 | 0% | 1,824 | 664 | -64% | 0 | 0 | — |
case-18 | fail→pass | 15,931 | 6,959 | -56% | 1 | 1 | 0% | 2,994 | 1,551 | -48% | 0 | 0 | — |
case-19 | fail→pass | 14,119 | 27,925 | +98% | 1 | 1 | 0% | 2,507 | 4,879 | +95% | 0 | 0 | — |
case-20 | pass→fail | 14,690 | 6,715 | -54% | 1 | 1 | 0% | 2,253 | 774 | -66% | 0 | 0 | — |
case-21 | pass→fail | 15,869 | 26,735 | +68% | 1 | 1 | 0% | 3,253 | 5,619 | +73% | 0 | 0 | — |
case-22 | pass→fail | 11,299 | 20,626 | +83% | 1 | 1 | 0% | 2,019 | 3,251 | +61% | 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, and 11 counted toward the lift figure. The other 11 produced results that are not comparable between the two arms, so they are excluded from the headline rather than averaged into it. The headline lift of 0 percentage points is the difference between those two pass rates over the 11 comparable cases. 6 cases got worse with the skill loaded, and they are 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.