Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Expert in building accessibility scanning tools, rule engines, document parsers, report generators, and audit automation. WCAG criterion mapping, severity scoring, CLI/GUI scanner architecture, CI/CD integration.
.claude/skills/community-access-accessibility-tool-builder/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-02 | ✗→✓ | ▲ Improved | 24% | 0% |
| case-01 | ✗→✓ | ▲ Improved | -2% | 0% |
| case-04 | ✗→✓ | ▲ Improved | 10% | 0% |
| case-05 | ✗→✓ | ▲ Improved | 35% | 0% |
| case-11 | ✗→✓ | ▲ Improved | 18% | 0% |
Shared instructions
You are an accessibility tool builder -- an expert in designing and building the scanning tools, rule engines, parsers, and report generators that power accessibility auditing workflows. You understand the architecture of tools like axe-core, pa11y, Accessibility Insights, and build equivalent tooling for desktop, documents, and custom domains.
Knowledge domains: Python Development, Web Scanning, Document Scanning
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-02 | fail→pass | 20,783 | 27,665 | +33% | 1 | 1 | 0% | 5,289 | 6,560 | +24% | 0 | 0 | — |
case-01 | fail→pass | 25,522 | 22,740 | -11% | 1 | 1 | 0% | 5,635 | 5,520 | -2% | 0 | 0 | — |
case-03 | pass→pass | 19,466 | 21,778 | +12% | 1 | 1 | 0% | 3,717 | 4,811 | +29% | 0 | 0 | — |
case-04 | fail→pass | 25,562 | 23,624 | -8% | 1 | 1 | 0% | 4,379 | 4,812 | +10% | 0 | 0 | — |
case-05 | fail→pass | 14,624 | 25,812 | +77% | 1 | 1 | 0% | 2,491 | 3,353 | +35% | 0 | 0 | — |
case-06 | fail→fail | 20,086 | 19,760 | -2% | 1 | 1 | 0% | 3,785 | 4,146 | +10% | 0 | 0 | — |
case-07 | pass→pass | 13,380 | 15,842 | +18% | 1 | 1 | 0% | 2,230 | 3,022 | +36% | 0 | 0 | — |
case-08 | pass→pass | 14,316 | 12,055 | -16% | 1 | 1 | 0% | 2,158 | 2,447 | +13% | 0 | 0 | — |
case-09 | pass→pass | 9,034 | 5,459 | -40% | 1 | 1 | 0% | 1,567 | 1,262 | -19% | 0 | 0 | — |
case-10 | pass→pass | 12,723 | 17,110 | +34% | 1 | 1 | 0% | 2,555 | 3,897 | +53% | 0 | 0 | — |
case-11 | fail→pass | 22,648 | 24,456 | +8% | 1 | 1 | 0% | 4,311 | 5,093 | +18% | 0 | 0 | — |
case-12 | pass→pass | 12,345 | 11,351 | -8% | 1 | 1 | 0% | 2,356 | 2,699 | +15% | 0 | 0 | — |
case-21 | pass→pass | 13,736 | 14,185 | +3% | 1 | 1 | 0% | 2,621 | 2,908 | +11% | 0 | 0 | — |
case-13 | pass→pass | 13,829 | 12,210 | -12% | 1 | 1 | 0% | 2,032 | 2,270 | +12% | 0 | 0 | — |
case-14 | fail→pass | 17,048 | 28,087 | +65% | 1 | 1 | 0% | 3,307 | 6,523 | +97% | 0 | 0 | — |
case-15 | fail→pass | 14,922 | 16,325 | +9% | 1 | 1 | 0% | 2,678 | 3,373 | +26% | 0 | 0 | — |
case-16 | pass→pass | 22,711 | 21,628 | -5% | 1 | 1 | 0% | 3,760 | 4,228 | +12% | 0 | 0 | — |
case-22 | pass→fail | 14,051 | 17,141 | +22% | 1 | 1 | 0% | 2,706 | 3,783 | +40% | 0 | 0 | — |
case-17 | pass→pass | 10,858 | 35,753 | +229% | 1 | 1 | 0% | 2,024 | 2,334 | +15% | 0 | 0 | — |
case-18 | pass→pass | 13,187 | 13,271 | +1% | 1 | 1 | 0% | 2,219 | 2,548 | +15% | 0 | 0 | — |
case-19 | fail→pass | 20,981 | 22,263 | +6% | 1 | 1 | 0% | 4,063 | 4,978 | +23% | 0 | 0 | — |
case-20 | pass→fail | 3,009 | 9,525 | +217% | 1 | 1 | 0% | 504 | 1,982 | +293% | 0 | 0 | — |
case-23 | pass→pass | 13,064 | 13,783 | +6% | 1 | 1 | 0% | 2,130 | 2,598 | +22% | 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. 23 cases were attempted. The headline lift of +26 percentage points is the difference between those two pass rates over the 23 comparable cases. 2 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.