Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →萌宠带货短视频全流程创作技能，基于COZE视频大模型API，支持萌宠剧情/好物测评/品种科普/宠品带货等全类型创作，覆盖爆款解析反推、知识库联动、互动优化、素材批量下载、全素材整合，实现从创意到带货引流型成品视频的自动化生成，适配15-30秒竖屏9:16主流平台规格。
.claude/skills/anbeime-pet-commerce-creator/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | 58% | 0% |
| case-02 | ✗→✓ | ▲ Improved | 63% | 0% |
| case-03 | ✗→✓ | ▲ Improved | 143% | 0% |
| case-04 | ✗→✓ | ▲ Improved | 59% | 0% |
| case-06 | ✗→✓ | ▲ Improved | 64% | 0% |
coze-workload-identity>=1.0.0 # COZE视频API调用 yt-dlp>=2024.1.0 # 视频下载（支持抖音/B站/YouTube等）
./downloads（自动创建）scripts/video_downloader.py下载视频及元数据scripts/coze_video_api.py处理素材上传、视频合成、状态轮询scripts/video_downloader.py download <URL>下载视频scripts/coze_video_api.py调用COZE API，使用coze-workload_identity的requestsbash python scripts/video_downloader.py download "https://www.bilibili.com/video/BV1xx411c7mD"
./downloads/视频标题.mp4./downloads/视频标题.info.json./downloads/视频标题.jpg| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 21,683 | 15,947 | -26% | 1 | 1 | 0% | 3,370 | 5,324 | +58% | 0 | 0 | — |
case-02 | fail→pass | 22,773 | 21,958 | -4% | 1 | 1 | 0% | 4,251 | 6,936 | +63% | 0 | 0 | — |
case-03 | fail→pass | 18,589 | 24,792 | +33% | 1 | 1 | 0% | 2,940 | 7,155 | +143% | 0 | 0 | — |
case-04 | fail→pass | 18,565 | 17,486 | -6% | 1 | 1 | 0% | 3,700 | 5,890 | +59% | 0 | 0 | — |
case-05 | fail→fail | 16,463 | 5,679 | -66% | 1 | 1 | 0% | 2,745 | 2,950 | +7% | 0 | 0 | — |
case-06 | fail→pass | 17,831 | 13,718 | -23% | 1 | 1 | 0% | 2,880 | 4,730 | +64% | 0 | 0 | — |
case-07 | fail→pass | 10,400 | 2,346 | -77% | 1 | 1 | 0% | 1,779 | 3,106 | +75% | 0 | 0 | — |
case-08 | fail→pass | 11,602 | 3,229 | -72% | 1 | 1 | 0% | 1,689 | 3,143 | +86% | 0 | 0 | — |
case-09 | fail→pass | 18,706 | 4,264 | -77% | 1 | 1 | 0% | 2,597 | 3,262 | +26% | 0 | 0 | — |
case-10 | fail→pass | 9,047 | 3,289 | -64% | 1 | 1 | 0% | 1,448 | 3,150 | +118% | 0 | 0 | — |
case-11 | fail→pass | 10,557 | 2,315 | -78% | 1 | 1 | 0% | 1,658 | 2,991 | +80% | 0 | 0 | — |
case-12 | fail→pass | 12,298 | 2,426 | -80% | 1 | 1 | 0% | 2,289 | 3,050 | +33% | 0 | 0 | — |
case-13 | fail→pass | 16,425 | 7,577 | -54% | 1 | 1 | 0% | 2,504 | 3,907 | +56% | 0 | 0 | — |
case-14 | fail→pass | 11,982 | 3,330 | -72% | 1 | 1 | 0% | 1,989 | 3,154 | +59% | 0 | 0 | — |
case-15 | fail→pass | 17,031 | 3,032 | -82% | 1 | 1 | 0% | 2,461 | 3,140 | +28% | 0 | 0 | — |
case-16 | fail→pass | 14,535 | 5,148 | -65% | 1 | 1 | 0% | 2,142 | 3,477 | +62% | 0 | 0 | — |
case-17 | fail→pass | 11,758 | 3,555 | -70% | 1 | 1 | 0% | 1,962 | 3,296 | +68% | 0 | 0 | — |
case-18 | fail→pass | 7,620 | 3,629 | -52% | 1 | 1 | 0% | 1,292 | 3,216 | +149% | 0 | 0 | — |
case-19 | fail→pass | 9,764 | 2,565 | -74% | 1 | 1 | 0% | 1,832 | 3,050 | +66% | 0 | 0 | — |
case-20 | fail→pass | 9,056 | 2,367 | -74% | 1 | 1 | 0% | 1,625 | 3,063 | +88% | 0 | 0 | — |
case-21 | pass→pass | 21,596 | 17,745 | -18% | 1 | 1 | 0% | 3,726 | 5,584 | +50% | 0 | 0 | — |
case-22 | pass→fail | 15,147 | 6,681 | -56% | 1 | 1 | 0% | 2,963 | 3,664 | +24% | 0 | 0 | — |
case-23 | pass→pass | 19,693 | 14,546 | -26% | 1 | 1 | 0% | 3,220 | 4,901 | +52% | 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 +78 percentage points is the difference between those two pass rates over the 22 comparable cases. 1 case got worse with the skill loaded, and it is 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.