Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →《三体》赛道AI视频创作工具,提供结构化的多智能体协作流程、素材生成与视频合成,涵盖选题深化、视觉设计、音频生成、视频制作全流程
.claude/skills/anbeime-three-body-video-creator/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-11 | ✗→✓ | ▲ Improved | -8% | 0% |
| case-18 | ✗→✓ | ▲ Improved | 148% | 0% |
| case-12 | ✗→✓ | ▲ Improved | 117% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 326% | 0% |
| case-08 | ✗→✓ | ▲ Improved | 103% | 0% |
opencv-python>=4.8.0 moviepy>=1.0.3 Pillow>=10.0.0
bash # 创建视频素材目录 mkdir -p ./raw-materials/{scenes,characters,audio} mkdir -p ./output/{draft,final}
scripts/voice_generator.py 生成角色配音--input <旁白数据JSON> --output ./output/audio/voice./output/audio/voice/scripts/sound_generator.py --type sound 生成音效--input <音效配置JSON> --output ./output/audio/sound_effects./output/audio/sound_effects/scripts/sound_generator.py --type music 生成背景音乐--input <音乐配置JSON> --output ./output/audio/background_music./output/audio/background_music/scripts/video_compositor.py 合成最终视频--images <图片目录> --audio <音频目录> --subtitles <字幕文件> --output <输出路径>./output/final/video.mp4scripts/quality_checker.py 检测:bash# 用户请求: "帮我创作一个《三体》威慑纪元的8分钟视频" # 智能体执行流程: 1. 读取 references/three-body-settings.md,提取核心设定 2. 按照步骤1-3完成前期筹备(时间线、剧情大纲、台词) 3. 按照步骤4-8完成视觉制作(场景、人物、分镜) 4. 按照步骤9-11完成音频与剪辑 5. 调用 scripts/quality_checker.py 检测技术指标 6. 完成步骤12-15的质量检测与优化
python# 检测视频技术指标 from scripts.quality_checker import check_video_quality result = check_video_quality( video_path="./output/final/video.mp4", expected_resolution=(1920, 1080), expected_duration=510, # 8分30秒 expected_fps=30 ) print(result) # 输出: {'resolution': True, 'duration': True, 'fps': True, 'overall': True}
bash# 用户请求: "为罗辑写一段移交威慑权的台词" # 智能体执行: 1. 读取 references/character-profiles.md,了解罗辑性格 2. 参考 references/three-body-settings.md,确保符合原著设定 3. 生成台词: "控制权移交给你们了。但记住,这是双刃剑——它在你们手中,也在我手中。" (标注:语气低沉平静,语速缓慢,略带疲惫但眼神锐利)
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-11 | fail→pass | 21,207 | 4,343 | -80% | 1 | 1 | 0% | 4,266 | 3,942 | -8% | 0 | 0 | — |
case-18 | fail→pass | 11,222 | 7,135 | -36% | 1 | 1 | 0% | 1,758 | 4,354 | +148% | 0 | 0 | — |
case-01 | fail→fail | 22,642 | 19,700 | -13% | 1 | 1 | 0% | 3,371 | 6,228 | +85% | 0 | 0 | — |
case-02 | fail→fail | 26,469 | 19,902 | -25% | 1 | 1 | 0% | 3,776 | 6,265 | +66% | 0 | 0 | — |
case-03 | fail→fail | 25,475 | 20,923 | -18% | 1 | 1 | 0% | 3,902 | 6,485 | +66% | 0 | 0 | — |
case-04 | pass→pass | 10,499 | 9,403 | -10% | 1 | 1 | 0% | 2,335 | 5,067 | +117% | 0 | 0 | — |
case-05 | pass→pass | 17,894 | 16,835 | -6% | 1 | 1 | 0% | 2,796 | 5,644 | +102% | 0 | 0 | — |
case-06 | pass→pass | 19,045 | 16,722 | -12% | 1 | 1 | 0% | 3,921 | 6,688 | +71% | 0 | 0 | — |
case-12 | fail→pass | 12,352 | 4,405 | -64% | 1 | 1 | 0% | 1,781 | 3,864 | +117% | 0 | 0 | — |
case-07 | fail→pass | 4,784 | 2,954 | -38% | 1 | 1 | 0% | 850 | 3,617 | +326% | 0 | 0 | — |
case-08 | fail→pass | 9,913 | 4,085 | -59% | 1 | 1 | 0% | 1,927 | 3,908 | +103% | 0 | 0 | — |
case-09 | fail→pass | 10,842 | 2,326 | -79% | 1 | 1 | 0% | 2,190 | 3,549 | +62% | 0 | 0 | — |
case-10 | fail→pass | 3,270 | 2,752 | -16% | 1 | 1 | 0% | 568 | 3,263 | +474% | 0 | 0 | — |
case-13 | fail→pass | 12,729 | 9,031 | -29% | 1 | 1 | 0% | 1,970 | 4,666 | +137% | 0 | 0 | — |
case-14 | fail→pass | 11,913 | 8,531 | -28% | 1 | 1 | 0% | 2,216 | 4,923 | +122% | 0 | 0 | — |
case-15 | fail→pass | 11,450 | 4,017 | -65% | 1 | 1 | 0% | 2,009 | 3,981 | +98% | 0 | 0 | — |
case-16 | fail→pass | 15,230 | 15,365 | +1% | 1 | 1 | 0% | 2,541 | 5,657 | +123% | 0 | 0 | — |
case-17 | pass→pass | 7,561 | 2,394 | -68% | 1 | 1 | 0% | 1,447 | 3,540 | +145% | 0 | 0 | — |
case-19 | pass→pass | 10,347 | 1,774 | -83% | 1 | 1 | 0% | 2,153 | 3,445 | +60% | 0 | 0 | — |
case-20 | pass→pass | 15,549 | 6,906 | -56% | 1 | 1 | 0% | 2,538 | 4,323 | +70% | 0 | 0 | — |
case-21 | fail→pass | 14,854 | 10,684 | -28% | 1 | 1 | 0% | 2,642 | 5,014 | +90% | 0 | 0 | — |
case-22 | fail→pass | 7,133 | 3,681 | -48% | 1 | 1 | 0% | 1,137 | 3,789 | +233% | 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 +59 percentage points is the difference between those two pass rates over the 22 comparable cases.
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.