Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Internationalization and RTL accessibility specialist. Audits dir attributes, BCP 47 lang tags, bidirectional text handling, mixed-direction forms, and icon mirroring in RTL. Ensures multilingual and RTL content is accessible to assistive technologies.
.claude/skills/community-access-i18n-accessibility-75cea7/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-05 | ✗→✓ | ▲ Improved | 122% | 0% |
| case-01 | ✓→✓ | ＝ Same ✓ | 24% | 0% |
| case-02 | ✓→✓ | ＝ Same ✓ | 24% | 0% |
| case-03 | ✓→✓ | ＝ Same ✓ | -13% | 0% |
| case-04 | ✓→✓ | ＝ Same ✓ | 81% | 0% |
You audit web content for internationalization-related accessibility issues — language identification, text direction, bidirectional content, and RTL layout correctness.
<html> must have valid BCP 47 lang attributelang attribute<html dir="rtl"> for RTL languagesdir="auto" for user-generated content<bdi> for inline bidirectional isolationmargin-inline-start not margin-left)email, url, tel) need dir="ltr" on input| Language | Tag | Direction | |----------|-----|-----------| | English | en | LTR | | Arabic | ar | RTL | | Hebrew | he | RTL | | Persian | fa | RTL | | Chinese | zh-Hans | LTR | | Japanese | ja | LTR |
<html lang>dir attributes and CSS logical properties| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | pass→pass | 4,191 | 3,391 | -19% | 1 | 1 | 0% | 736 | 910 | +24% | 0 | 0 | — |
case-02 | pass→pass | 5,028 | 5,037 | +0% | 1 | 1 | 0% | 925 | 1,146 | +24% | 0 | 0 | — |
case-03 | pass→pass | 7,831 | 4,329 | -45% | 1 | 1 | 0% | 1,270 | 1,102 | -13% | 0 | 0 | — |
case-04 | pass→pass | 3,656 | 4,125 | +13% | 1 | 1 | 0% | 585 | 1,060 | +81% | 0 | 0 | — |
case-05 | fail→pass | 2,610 | 3,876 | +49% | 1 | 1 | 0% | 426 | 945 | +122% | 0 | 0 | — |
case-06 | pass→pass | 4,699 | 5,026 | +7% | 1 | 1 | 0% | 732 | 1,138 | +55% | 0 | 0 | — |
case-07 | pass→pass | 2,614 | 2,618 | +0% | 1 | 1 | 0% | 500 | 785 | +57% | 0 | 0 | — |
case-08 | pass→pass | 3,180 | 2,515 | -21% | 1 | 1 | 0% | 540 | 775 | +44% | 0 | 0 | — |
case-09 | pass→pass | 8,292 | 7,907 | -5% | 1 | 1 | 0% | 1,292 | 1,557 | +21% | 0 | 0 | — |
case-10 | pass→pass | 10,554 | 9,012 | -15% | 1 | 1 | 0% | 1,689 | 1,865 | +10% | 0 | 0 | — |
case-11 | pass→pass | 7,745 | 4,085 | -47% | 1 | 1 | 0% | 1,231 | 978 | -21% | 0 | 0 | — |
case-12 | pass→pass | 10,020 | 6,802 | -32% | 1 | 1 | 0% | 1,759 | 1,441 | -18% | 0 | 0 | — |
case-13 | pass→pass | 6,530 | 6,027 | -8% | 1 | 1 | 0% | 1,031 | 1,295 | +26% | 0 | 0 | — |
case-14 | pass→pass | 3,033 | 2,356 | -22% | 1 | 1 | 0% | 519 | 679 | +31% | 0 | 0 | — |
case-15 | pass→pass | 2,812 | 3,180 | +13% | 1 | 1 | 0% | 460 | 680 | +48% | 0 | 0 | — |
case-16 | pass→pass | 2,531 | 2,698 | +7% | 1 | 1 | 0% | 347 | 686 | +98% | 0 | 0 | — |
case-17 | pass→pass | 3,858 | 2,262 | -41% | 1 | 1 | 0% | 573 | 647 | +13% | 0 | 0 | — |
case-18 | pass→pass | 5,004 | 4,322 | -14% | 1 | 1 | 0% | 824 | 964 | +17% | 0 | 0 | — |
case-19 | pass→pass | 3,562 | 3,516 | -1% | 1 | 1 | 0% | 610 | 884 | +45% | 0 | 0 | — |
case-20 | pass→pass | 3,026 | 3,254 | +8% | 1 | 1 | 0% | 562 | 870 | +55% | 0 | 0 | — |
case-21 | pass→pass | 11,767 | 10,971 | -7% | 1 | 1 | 0% | 2,176 | 2,238 | +3% | 0 | 0 | — |
case-22 | pass→pass | 5,327 | 5,818 | +9% | 1 | 1 | 0% | 982 | 1,323 | +35% | 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 +5 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.