Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →拟真 X 推文卡片 + 互动数据 (likes/reposts/views), 适配视频叠加或图卡分享
.claude/skills/nexu-io-social-x-post-card/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-04 | ✗→✓ | ▲ Improved | 25% | 0% |
| case-15 | ✗→✓ | ▲ Improved | 18% | 0% |
| case-18 | ✗→✓ | ▲ Improved | 33% | 0% |
| case-20 | ✗→✓ | ▲ Improved | 21% | 0% |
| case-21 | ✓→✗ | ▼ Worse | 82% | 0% |
【模板: X (Twitter) 帖子卡】 【意图】把一段推文内容 (或用户的金句) 渲染成一张拟真度极高的 X 帖子卡片, 用于视频叠加、推特发图、知识沉淀。Inspired by hyperframes x-post。
【画布】1280×720 或 1080×1080, 暗背景 #0f1419 或亮背景 #ffffff (按 X 主题); 卡片居中, 阴影柔和。
【卡片结构】
#2f3336 (dark) / #eff3f4 (light), 内边距 16px。@username + verified 蓝勾 + 时间 (mono, 12px, 灰)。#1d9bf0; hashtag 同色; mention 同色; 段落间空 0.6em。【字体】
Chirp (X 的字体) → fallback Inter 或 Segoe UI。Noto Sans SC / PingFang SC。【设计细节】
#fff, text #0f1419, secondary #536471, border #eff3f4, accent #1d9bf0。#000, text #e7e9ea, secondary #71767b, border #2f3336, accent #1d9bf0。radial-gradient(...) 增加视频叠加的可读性。| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 20,670 | 27,176 | +31% | 1 | 1 | 0% | 4,880 | 6,882 | +41% | 0 | 0 | — |
case-02 | fail→fail | 26,445 | 27,004 | +2% | 1 | 1 | 0% | 6,191 | 6,872 | +11% | 0 | 0 | — |
case-03 | fail→fail | 25,190 | 24,320 | -3% | 1 | 1 | 0% | 5,754 | 6,318 | +10% | 0 | 0 | — |
case-04 | fail→pass | 26,579 | 26,131 | -2% | 1 | 1 | 0% | 5,492 | 6,862 | +25% | 0 | 0 | — |
case-05 | fail→fail | 28,440 | 27,191 | -4% | 1 | 1 | 0% | 6,198 | 6,879 | +11% | 0 | 0 | — |
case-06 | fail→fail | 25,876 | 25,354 | -2% | 1 | 1 | 0% | 6,175 | 6,856 | +11% | 0 | 0 | — |
case-07 | fail→fail | 22,724 | 29,078 | +28% | 1 | 1 | 0% | 5,038 | 6,865 | +36% | 0 | 0 | — |
case-08 | pass→pass | 18,369 | 27,460 | +49% | 1 | 1 | 0% | 4,136 | 6,857 | +66% | 0 | 0 | — |
case-09 | fail→fail | 29,213 | 28,982 | -1% | 1 | 1 | 0% | 6,180 | 6,798 | +10% | 0 | 0 | — |
case-10 | fail→fail | 27,865 | 26,599 | -5% | 1 | 1 | 0% | 6,172 | 6,481 | +5% | 0 | 0 | — |
case-11 | fail→fail | 25,419 | 25,038 | -1% | 1 | 1 | 0% | 5,621 | 6,864 | +22% | 0 | 0 | — |
case-12 | fail→fail | 5,036 | 25,584 | +408% | 1 | 1 | 0% | 1,004 | 6,853 | +583% | 0 | 0 | — |
case-13 | fail→fail | 23,632 | 25,687 | +9% | 1 | 1 | 0% | 5,159 | 6,861 | +33% | 0 | 0 | — |
case-14 | fail→fail | 22,615 | 27,321 | +21% | 1 | 1 | 0% | 5,142 | 6,853 | +33% | 0 | 0 | — |
case-15 | fail→pass | 25,791 | 27,326 | +6% | 1 | 1 | 0% | 5,502 | 6,502 | +18% | 0 | 0 | — |
case-16 | fail→fail | 26,349 | 26,933 | +2% | 1 | 1 | 0% | 5,712 | 6,884 | +21% | 0 | 0 | — |
case-17 | fail→fail | 28,201 | 29,145 | +3% | 1 | 1 | 0% | 6,170 | 6,851 | +11% | 0 | 0 | — |
case-18 | fail→pass | 21,768 | 28,225 | +30% | 1 | 1 | 0% | 5,157 | 6,855 | +33% | 0 | 0 | — |
case-19 | fail→fail | 24,938 | 26,275 | +5% | 1 | 1 | 0% | 5,675 | 6,726 | +19% | 0 | 0 | — |
case-20 | fail→pass | 24,462 | 25,091 | +3% | 1 | 1 | 0% | 5,690 | 6,871 | +21% | 0 | 0 | — |
case-21 | pass→fail | 15,178 | 25,729 | +70% | 1 | 1 | 0% | 3,773 | 6,875 | +82% | 0 | 0 | — |
case-22 | pass→fail | 13,942 | 25,531 | +83% | 1 | 1 | 0% | 3,245 | 6,861 | +111% | 0 | 0 | — |
case-23 | pass→fail | 12,407 | 27,784 | +124% | 1 | 1 | 0% | 2,292 | 6,853 | +199% | 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 +4 percentage points is the difference between those two pass rates over the 23 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.