Install any skill in seconds. Free to start, no credit card required.
Get Started Free →提供基于 FFmpeg 和 ImageMagick 的多媒体处理能力,支持视频和图像的格式转换、分辨率调整、压缩等操作
.claude/skills/anbeime-media-processor/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-04 | ✗→✓ | ▲ Improved | 0% | 0% |
| case-03 | ✗→✓ | ▲ Improved | 80% | 0% |
| case-05 | ✗→✓ | ▲ Improved | 285% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 830% | 0% |
| case-08 | ✗→✓ | ▲ Improved | 64% | 0% |
bash apt-get update && apt-get install -y ffmpeg imagemagick
scripts/ 中的视频处理脚本scripts/ 中的图像处理脚本scripts/video_convert.pyscripts/video_compress.pyscripts/video_scale.pyscripts/image_convert.pyscripts/image_scale.pyscripts/image_compress.py*.avi、images/*.png)./ 开头)python# 将 video.avi 转换为 video.mp4 python scripts/video_convert.py \ --input ./video.avi \ --output ./video.mp4 \ --format mp4
python# 将所有 PNG 转换为 JPG for file in ./images/*.png; do output="${file%.png}.jpg" python scripts/image_convert.py \ --input "$file" \ --output "$output" \ --format jpg \ --quality 85 done
python# 先缩放再压缩 python scripts/video_scale.py \ --input ./input.mp4 \ --output ./temp.mp4 \ --width 1280 \ --height 720 python scripts/video_compress.py \ --input ./temp.mp4 \ --output ./output.mp4 \ --bitrate 2M
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-04 | fail→pass | 11,231 | 4,713 | -58% | 1 | 1 | 0% | 2,196 | 2,198 | +0% | 0 | 0 | — |
case-01 | fail→fail | 4,834 | 5,498 | +14% | 1 | 1 | 0% | 951 | 2,395 | +152% | 0 | 0 | — |
case-02 | fail→fail | 9,930 | 6,026 | -39% | 1 | 1 | 0% | 2,088 | 1,693 | -19% | 0 | 0 | — |
case-03 | fail→pass | 6,384 | 5,385 | -16% | 1 | 1 | 0% | 1,382 | 2,481 | +80% | 0 | 0 | — |
case-05 | fail→pass | 3,202 | 2,133 | -33% | 1 | 1 | 0% | 426 | 1,640 | +285% | 0 | 0 | — |
case-06 | pass→pass | 9,236 | 2,244 | -76% | 1 | 1 | 0% | 1,800 | 1,699 | -6% | 0 | 0 | — |
case-07 | fail→pass | 5,471 | 2,209 | -60% | 1 | 1 | 0% | 187 | 1,739 | +830% | 0 | 0 | — |
case-08 | fail→pass | 7,091 | 3,374 | -52% | 1 | 1 | 0% | 1,048 | 1,714 | +64% | 0 | 0 | — |
case-09 | fail→pass | 7,689 | 2,141 | -72% | 1 | 1 | 0% | 1,321 | 1,578 | +19% | 0 | 0 | — |
case-10 | fail→pass | 4,777 | 2,224 | -53% | 1 | 1 | 0% | 789 | 1,528 | +94% | 0 | 0 | — |
case-11 | pass→pass | 5,518 | 2,591 | -53% | 1 | 1 | 0% | 969 | 1,722 | +78% | 0 | 0 | — |
case-12 | pass→pass | 5,370 | 3,107 | -42% | 1 | 1 | 0% | 909 | 1,795 | +97% | 0 | 0 | — |
case-13 | pass→pass | 5,846 | 2,619 | -55% | 1 | 1 | 0% | 1,033 | 1,623 | +57% | 0 | 0 | — |
case-14 | pass→pass | 11,536 | 2,685 | -77% | 1 | 1 | 0% | 1,617 | 1,878 | +16% | 0 | 0 | — |
case-15 | fail→pass | 8,858 | 1,450 | -84% | 1 | 1 | 0% | 1,654 | 1,538 | -7% | 0 | 0 | — |
case-16 | pass→pass | 7,855 | 2,282 | -71% | 1 | 1 | 0% | 1,138 | 1,699 | +49% | 0 | 0 | — |
case-17 | fail→pass | 5,277 | 1,323 | -75% | 1 | 1 | 0% | 1,093 | 1,462 | +34% | 0 | 0 | — |
case-18 | pass→pass | 9,553 | 3,080 | -68% | 1 | 1 | 0% | 1,450 | 1,802 | +24% | 0 | 0 | — |
case-19 | pass→pass | 7,249 | 1,834 | -75% | 1 | 1 | 0% | 1,101 | 1,617 | +47% | 0 | 0 | — |
case-20 | fail→pass | 4,623 | 5,274 | +14% | 1 | 1 | 0% | 949 | 2,255 | +138% | 0 | 0 | — |
case-21 | fail→fail | 9,072 | 13,753 | +52% | 1 | 1 | 0% | 1,320 | 3,442 | +161% | 0 | 0 | — |
case-22 | fail→fail | 9,715 | 9,981 | +3% | 1 | 1 | 0% | 1,739 | 2,879 | +66% | 0 | 0 | — |
case-23 | fail→fail | 7,103 | 7,030 | -1% | 1 | 1 | 0% | 1,604 | 2,073 | +29% | 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, and 22 counted toward the lift figure. The other 1 produced results that are not comparable between the two arms, so they are excluded from the headline rather than averaged into it. The headline lift of +43 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.