Install any skill in seconds. Free to start, no credit card required.
Get Started Free →数字故障 / 像散偏移 / 数据腐败标题, 适合视频转场 / cyberpunk hero
.claude/skills/nexu-io-frame-glitch-title/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-04 | ✗→✓ | ▲ Improved | 20% | 0% |
| case-11 | ✗→✓ | ▲ Improved | 39% | 0% |
| case-14 | ✗→✓ | ▲ Improved | 11% | 0% |
| case-07 | ✓→✗ | ▼ Worse | 67% | 0% |
| case-21 | ✓→✗ | ▼ Worse | 69% | 0% |
【模板: 故障艺术标题帧 (Glitch Title)】 【意图】单帧 hero / 视频转场 / cyberpunk 风格标题。Inspired by hyperframes glitch。
【画布】1920×1080, 背景 #070708 近黑或 CRT 暗灰 #0d0e10; 加 56px 网格 (透明 5%) + scanlines 横线 (透明 8%, 2px 间隔)。
【主标题】
Space Grotesk Bold / Inter Tight Black / JetBrains Mono Bold。#f5f5f7; 后面套 2 层伪影:#00f0ff translate(-3px, 1px)。#ff2bd6 translate(3px, -1px)。@keyframes 随机 translateX -10px → 10px, 持续 80-160ms, 错峰播放, 营造 "data corruption" 像散。filter: url(#displacementFilter) 或简单 CSS 平移。【附加层】
>> SIGNAL_LOST · CH-04 · 14:32:08。 ̶▒̶ 字符替换 (假乱码)。█▓▒░ ASCII 噪点 chunks。background-image: url("data:image/svg+xml,...turbulence..."), opacity 6%, mix-blend-mode overlay。【SVG 滤镜 (可选)】
<filter id="rgbShift"> 用 feColorMatrix + feOffset + feMerge 把 R/G/B 三通道偏移; 整层 filter: url(#rgbShift) 在故障瞬间应用。【设计细节】
Space Grotesk 或 JetBrains Mono Bold; 中文 Noto Sans Mono CJK SC 或 Noto Sans SC Bold。@keyframes, 可被 prefers-reduced-motion 关闭 (退回静态 chromatic split)。| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-04 | fail→pass | 24,948 | 27,394 | +10% | 1 | 1 | 0% | 5,753 | 6,885 | +20% | 0 | 0 | — |
case-05 | fail→fail | 20,321 | 25,224 | +24% | 1 | 1 | 0% | 4,751 | 6,486 | +37% | 0 | 0 | — |
case-18 | fail→fail | 20,867 | 25,019 | +20% | 1 | 1 | 0% | 4,919 | 6,847 | +39% | 0 | 0 | — |
case-02 | fail→fail | 31,101 | 27,925 | -10% | 1 | 1 | 0% | 6,212 | 6,881 | +11% | 0 | 0 | — |
case-03 | fail→fail | 27,766 | 24,412 | -12% | 1 | 1 | 0% | 6,210 | 6,880 | +11% | 0 | 0 | — |
case-01 | fail→fail | 27,689 | 26,137 | -6% | 1 | 1 | 0% | 6,219 | 6,887 | +11% | 0 | 0 | — |
case-06 | fail→fail | 23,459 | 27,681 | +18% | 1 | 1 | 0% | 5,506 | 6,861 | +25% | 0 | 0 | — |
case-07 | pass→fail | 18,518 | 28,951 | +56% | 1 | 1 | 0% | 4,093 | 6,848 | +67% | 0 | 0 | — |
case-08 | fail→fail | 26,696 | 24,789 | -7% | 1 | 1 | 0% | 6,188 | 6,858 | +11% | 0 | 0 | — |
case-09 | fail→fail | 28,759 | 26,979 | -6% | 1 | 1 | 0% | 6,185 | 6,854 | +11% | 0 | 0 | — |
case-10 | fail→fail | 28,411 | 26,388 | -7% | 1 | 1 | 0% | 6,178 | 6,847 | +11% | 0 | 0 | — |
case-11 | fail→pass | 21,976 | 25,569 | +16% | 1 | 1 | 0% | 4,965 | 6,894 | +39% | 0 | 0 | — |
case-12 | fail→fail | 24,271 | 29,162 | +20% | 1 | 1 | 0% | 5,620 | 6,847 | +22% | 0 | 0 | — |
case-13 | fail→fail | 19,078 | 26,774 | +40% | 1 | 1 | 0% | 4,177 | 6,860 | +64% | 0 | 0 | — |
case-14 | fail→pass | 28,238 | 26,797 | -5% | 1 | 1 | 0% | 6,181 | 6,849 | +11% | 0 | 0 | — |
case-15 | fail→fail | 26,834 | 27,479 | +2% | 1 | 1 | 0% | 6,182 | 6,851 | +11% | 0 | 0 | — |
case-16 | pass→pass | 24,533 | 28,197 | +15% | 1 | 1 | 0% | 5,317 | 6,853 | +29% | 0 | 0 | — |
case-17 | fail→fail | 25,372 | 26,489 | +4% | 1 | 1 | 0% | 5,269 | 6,849 | +30% | 0 | 0 | — |
case-19 | fail→fail | 26,934 | 27,150 | +1% | 1 | 1 | 0% | 6,189 | 6,858 | +11% | 0 | 0 | — |
case-20 | pass→pass | 11,627 | 16,998 | +46% | 1 | 1 | 0% | 2,747 | 4,473 | +63% | 0 | 0 | — |
case-21 | pass→fail | 17,231 | 28,486 | +65% | 1 | 1 | 0% | 4,045 | 6,837 | +69% | 0 | 0 | — |
case-22 | pass→fail | 12,456 | 27,248 | +119% | 1 | 1 | 0% | 2,364 | 6,839 | +189% | 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 0 percentage points is the difference between those two pass rates over the 22 comparable cases. 4 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.