Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →胶片漏光 + 颗粒噪点 + 16:9 letterbox + 衬线大字, 电影感开场 / 章节卡
.claude/skills/nexu-io-frame-light-leak-cinema/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-02 | ✗→✓ | ▲ Improved | 13% | 0% |
| case-06 | ✗→✓ | ▲ Improved | 27% | 0% |
| case-12 | ✗→✓ | ▲ Improved | 19% | 0% |
| case-13 | ✗→✓ | ▲ Improved | 36% | 0% |
| case-19 | ✗→✓ | ▲ Improved | 4% | 0% |
【模板: 胶片漏光电影帧】 【意图】纪录片 / 个人短片 / 视频章节卡的开场单帧 —— 暖橙漏光 + 35mm 颗粒 + 衬线大字, 古典胶片质感。Inspired by hyperframes light-leak。
【画布】
#000)。【背景】
#1a0d08 / 墨绿 #0a1410 / 蓝紫 #0d0e1a) 或场景描绘 (CSS gradient 模拟天空 / 室内 / 室外)。radial-gradient(ellipse at top right, #ffb547 0%, transparent 50%) + 1 个底部 linear-gradient(to top, #d97757 0%, transparent 30%); 颜色取暖橙 / 桃 / 玫红 / 暗黄, 不要冷蓝。mix-blend-mode: overlay; 也可用 background-image: url("data:image/svg+xml,...feTurbulence...")。feDisplacementMap 模拟胶片摆动 (慎用)。【文字】
#f5e9d6 或 cream。【可选附加】
box-shadow 多重 inset 或多个 <div>)。【设计细节】
Noto Serif SC italic 不存在 → 用 Noto Serif SC regular + 字距加大。prefers-reduced-motion 关动效。| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-09 | pass→fail | 29,334 | 28,702 | -2% | 1 | 1 | 0% | 6,166 | 6,955 | +13% | 0 | 0 | — |
case-07 | fail→fail | 25,942 | 27,708 | +7% | 1 | 1 | 0% | 6,189 | 6,977 | +13% | 0 | 0 | — |
case-08 | pass→pass | 22,927 | 27,726 | +21% | 1 | 1 | 0% | 4,905 | 6,955 | +42% | 0 | 0 | — |
case-01 | fail→fail | 29,967 | 29,338 | -2% | 1 | 1 | 0% | 6,209 | 6,997 | +13% | 0 | 0 | — |
case-02 | fail→pass | 29,459 | 27,824 | -6% | 1 | 1 | 0% | 6,208 | 6,996 | +13% | 0 | 0 | — |
case-03 | fail→fail | 28,729 | 27,584 | -4% | 1 | 1 | 0% | 6,203 | 6,991 | +13% | 0 | 0 | — |
case-04 | pass→pass | 28,189 | 26,521 | -6% | 1 | 1 | 0% | 6,189 | 6,977 | +13% | 0 | 0 | — |
case-05 | fail→fail | 30,875 | 26,433 | -14% | 1 | 1 | 0% | 6,178 | 6,520 | +6% | 0 | 0 | — |
case-06 | fail→pass | 26,611 | 27,444 | +3% | 1 | 1 | 0% | 5,482 | 6,971 | +27% | 0 | 0 | — |
case-10 | fail→fail | 19,278 | 29,142 | +51% | 1 | 1 | 0% | 4,128 | 6,956 | +69% | 0 | 0 | — |
case-11 | pass→pass | 28,479 | 25,875 | -9% | 1 | 1 | 0% | 6,173 | 6,565 | +6% | 0 | 0 | — |
case-12 | fail→pass | 30,008 | 27,222 | -9% | 1 | 1 | 0% | 5,833 | 6,953 | +19% | 0 | 0 | — |
case-13 | fail→pass | 23,049 | 27,700 | +20% | 1 | 1 | 0% | 5,121 | 6,960 | +36% | 0 | 0 | — |
case-14 | pass→pass | 27,968 | 25,268 | -10% | 1 | 1 | 0% | 6,167 | 6,170 | +0% | 0 | 0 | — |
case-15 | fail→fail | 31,685 | 29,580 | -7% | 1 | 1 | 0% | 6,168 | 6,956 | +13% | 0 | 0 | — |
case-16 | pass→pass | 30,108 | 28,195 | -6% | 1 | 1 | 0% | 6,165 | 6,483 | +5% | 0 | 0 | — |
case-17 | fail→fail | 19,794 | 28,604 | +45% | 1 | 1 | 0% | 4,713 | 6,898 | +46% | 0 | 0 | — |
case-18 | pass→pass | 20,331 | 27,489 | +35% | 1 | 1 | 0% | 4,472 | 6,949 | +55% | 0 | 0 | — |
case-19 | fail→pass | 27,307 | 22,327 | -18% | 1 | 1 | 0% | 5,738 | 5,963 | +4% | 0 | 0 | — |
case-20 | pass→fail | 21,351 | 28,880 | +35% | 1 | 1 | 0% | 5,027 | 6,959 | +38% | 0 | 0 | — |
case-21 | pass→fail | 11,962 | 25,445 | +113% | 1 | 1 | 0% | 2,124 | 6,973 | +228% | 0 | 0 | — |
case-22 | fail→fail | 32,704 | 28,486 | -13% | 1 | 1 | 0% | 6,170 | 6,958 | +13% | 0 | 0 | — |
case-23 | fail→pass | 30,320 | 29,120 | -4% | 1 | 1 | 0% | 6,177 | 6,965 | +13% | 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 +13 percentage points is the difference between those two pass rates over the 23 comparable cases. 5 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.