Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →A self-contained WebGL2 hero: thin-film interference over a crushed-foil surface whose palette shifts with the viewing angle; move the cursor to tilt the film.
.claude/skills/nexu-io-webgl-holographic-foil/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-21 | ✗→✓ | ▲ Improved | 146% | 0% |
| case-03 | ✓→✗ | ▼ Worse | 129% | 0% |
| case-04 | ✓→✗ | ▼ Worse | 103% | 0% |
| case-12 | ✓→✗ | ▼ Worse | 34% | 0% |
| case-15 | ✓→✗ | ▼ Worse | 3% | 0% |
Produce a single self-contained index.html — A self-contained WebGL2 hero: thin-film interference over a crushed-foil surface whose palette shifts with the viewing angle; move the cursor to tilt the film.
OpenDesign detects getContext('webgl2') / heavy WebGL and renders this file in powered preview (a cross-origin-isolated iframe). The full GPU + scroll pipeline runs; no opaque-sandbox workarounds are needed.
webgl-holographic-foil/
├── SKILL.md ← you're reading this
├── example.html ← the complete, working artifact (READ FIRST)
└── (assets, if any)Keep any bundled LICENSE and on-screen credit intact. Replace imagery only with license-clean assets (original / AI, Lummi.ai, Unsplash/Pexels — never scraped imagery).
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 43,498 | 48,324 | +11% | 1 | 1 | 0% | 8,238 | 8,465 | +3% | 0 | 0 | — |
case-02 | fail→fail | 39,557 | 100,369 | +154% | 1 | 1 | 0% | 8,231 | 8,458 | +3% | 0 | 0 | — |
case-03 | pass→fail | 18,904 | 42,699 | +126% | 1 | 1 | 0% | 3,705 | 8,474 | +129% | 0 | 0 | — |
case-04 | pass→fail | 19,322 | 79,618 | +312% | 1 | 1 | 0% | 4,176 | 8,458 | +103% | 0 | 0 | — |
case-05 | fail→fail | 21,122 | 36,059 | +71% | 1 | 1 | 0% | 3,420 | 8,451 | +147% | 0 | 0 | — |
case-06 | fail→fail | 39,369 | 52,396 | +33% | 1 | 1 | 0% | 8,229 | 8,456 | +3% | 0 | 0 | — |
case-07 | fail→fail | 41,623 | 54,635 | +31% | 1 | 1 | 0% | 8,229 | 8,456 | +3% | 0 | 0 | — |
case-08 | fail→fail | 42,057 | 43,147 | +3% | 1 | 1 | 0% | 8,228 | 8,455 | +3% | 0 | 0 | — |
case-09 | fail→fail | 47,841 | 40,399 | -16% | 1 | 1 | 0% | 8,232 | 8,459 | +3% | 0 | 0 | — |
case-10 | fail→fail | 38,946 | 46,098 | +18% | 1 | 1 | 0% | 8,244 | 8,471 | +3% | 0 | 0 | — |
case-11 | fail→fail | 55,619 | 52,928 | -5% | 1 | 1 | 0% | 8,221 | 8,448 | +3% | 0 | 0 | — |
case-12 | pass→fail | 44,613 | 52,988 | +19% | 1 | 1 | 0% | 6,330 | 8,452 | +34% | 0 | 0 | — |
case-13 | fail→fail | 40,968 | 45,460 | +11% | 1 | 1 | 0% | 8,219 | 8,446 | +3% | 0 | 0 | — |
case-14 | fail→fail | 43,145 | 42,286 | -2% | 1 | 1 | 0% | 8,233 | 8,460 | +3% | 0 | 0 | — |
case-15 | pass→fail | 37,587 | 39,727 | +6% | 1 | 1 | 0% | 8,233 | 8,460 | +3% | 0 | 0 | — |
case-16 | pass→fail | 40,194 | 41,916 | +4% | 1 | 1 | 0% | 8,221 | 8,448 | +3% | 0 | 0 | — |
case-17 | pass→pass | 40,985 | 45,599 | +11% | 1 | 1 | 0% | 8,220 | 8,447 | +3% | 0 | 0 | — |
case-18 | fail→fail | 52,134 | 52,274 | +0% | 1 | 1 | 0% | 8,221 | 8,448 | +3% | 0 | 0 | — |
case-19 | pass→fail | 26,029 | 55,086 | +112% | 1 | 1 | 0% | 4,989 | 8,449 | +69% | 0 | 0 | — |
case-20 | fail→fail | 38,215 | 41,795 | +9% | 1 | 1 | 0% | 8,222 | 8,449 | +3% | 0 | 0 | — |
case-21 | fail→pass | 21,268 | 48,054 | +126% | 1 | 1 | 0% | 3,149 | 7,748 | +146% | 0 | 0 | — |
case-22 | fail→fail | 40,252 | 42,222 | +5% | 1 | 1 | 0% | 8,223 | 8,450 | +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 -64 percentage points is the difference between those two pass rates over the 22 comparable cases. 9 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.