Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add yehyakin/hermes-skills --skill ecommerce-visual-clip-scanninggit clone --depth 1 https://github.com/yehyakin/hermes-skillsWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/yehyakin/hermes-skills/ecommerce-visual-clip-scanning)<a href="https://agentmods.dev/skills/yehyakin/hermes-skills/ecommerce-visual-clip-scanning"><img src="https://agentmods.dev/badge/skills/yehyakin/hermes-skills/ecommerce-visual-clip-scanning/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/yehyakin/hermes-skills/ecommerce-visual-clip-scanning"><img src="https://agentmods.dev/badge/skills/yehyakin/hermes-skills/ecommerce-visual-clip-scanning.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00062 | $0.01581 |
| Opus 5 | $0.00031 | $0.00790 |
| Sonnet 5 | $0.00012 | $0.00316 |
| Haiku 4.5 | $0.00006 | $0.00158 |
Grade A, and why
ecommerce-visual-clip-scanning scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 11d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 141 lines — stays where its author put it; the contents beside it link to each section on GitHub.
电商视频视觉切片扫描法
通过画面抽帧 + 视觉AI分析识别直播/竞品视频中的带货高光片段。替代 Whisper 转写方法,更可靠(主播说到产品时镜头可能没对着产品)。
适用场景
- 直播回放:主播说到"这件外套"但镜头可能在拍书架/空镜
- 没有字幕的竞品视频
- 验证 transcript/关键词方法找到的候选片段是否真有产品展示
完整流程
Step 1: 确认视频路径和规格
# 找视频文件
find /Users/yehya -type f -name "*.mp4" -size +100M 2>/dev/null | grep -v Library
# 确认时长、分辨率、编码
ffprobe -v quiet -print_format json -show_format -show_streams "VIDEO.mp4" 2>&1 | grep -E '"duration"|"width"|"height"|"codec_name"'
Step 2: 全局采样抽帧(每5分钟1帧)
INPUT="/path/to/video.mp4"
OUTPUT_DIR="/tmp/scan"
DURATION=$(ffprobe -v quiet -show_entries format=duration -of csv=p=0 "$INPUT")
mkdir -p "$OUTPUT_DIR"
for i in $(seq 0 300 $DURATION); do
ffmpeg -ss $i -i "$INPUT" -vframes 1 -q:v 2 "$OUTPUT_DIR/frame_$(printf "%04d" $i).jpg" -y 2>/dev/null
echo -n "."
done
echo ""
echo "完成: $(ls $OUTPUT_DIR/frame_*.jpg | wc -l) 帧"
经验:5分钟间隔是平衡精度和工作量的最优值。3.6小时视频 → 44帧。
Step 3: 批量视觉AI分析
每批5张并行分析(避免单批过多延迟):
每批次问题模板:
"主播在展示产品吗?画面里有没有拿着衣服/鞋子/配饰?简述画面内容。"
判断标准:
✅ 有产品:主播手持/举起服装/鞋/包
⚠️ 不确定:主播在穿/戴产品(可接受)
❌ 无产品:主播只说话/打电话/看手机/空镜
Step 4: 定位时间 → 二次精确验证
初筛发现的时间点,在前后±30秒做精确验证:
# 精细验证帧
for ts in 2900 2930 2960 2990; do
ffmpeg -ss $ts -i "$INPUT" -vframes 1 -q:v 2 "$OUTPUT_DIR/verify_$(printf '%06d' $ts).jpg" -y 2>/dev/null
done
教训:5分钟采样会漏掉短于5分钟的展示段(本案例:毛衣展示在54分钟,5分钟采样跳过了)。 解决方案:在发现产品类型后,对该时段做密集扫描(每30秒1帧)。
Step 5: 切片段
INPUT="/path/to/video.mp4"
OUTPUT="/path/to/clips"
# 时间点(秒)→ 时:分:秒
ffmpeg -ss 3240 -i "$INPUT" -t 25 "$OUTPUT/精彩片段.mp4" -y 2>/dev/null
# 3240秒 = 54分钟
Step 6: 验证片段画面质量
必须步骤:从切好的片段中抽帧再次确认,不能假设切的是对的。
# 从片段第5秒抽1帧验证
ffmpeg -ss 5 -i "$OUTPUT/clip.mp4" -vframes 1 "$OUTPUT/vc.jpg" -y 2>/dev/null
# vision_analyze 用 file:// 前缀
vision_analyze(image_url="file:///path/to/vc.jpg", question="主播有没有在展示产品?手持衣服/皮鞋吗?")
Step 7: 交付或后续处理
- 直接交付 MP4 片段(无字幕)
- 或用 mcp-video 工具链烧字幕、加字幕
- 片段规格:20-25秒,H264,竖屏 1088x1920
关键教训
❌ 不要只信字幕/Transcript
教训来源(2026/04/25): 用 transcript 文字选了5个片段,字面上都是"带货话术",但验画面后发现:
- 片段5:主播站窗前说话,背景书架,看不到产品
- 片段7:主播坐椅子上说话,全程没拿产品
- 片段10:主播根本没入镜,3帧全是空镜
What ships with it
1 file beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 11d ago First seen · 141 lines · 62 tokens per session scan A 402059a319ed
ecommerce-visual-clip-scanning is a skill published in the GitHub repository yehyakin/hermes-skills (9 stars, last pushed 3mo ago), licensed MIT. It adds 62 tokens to every session and 1,581 once invoked, about $0.0003 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other skills, from other repositories
ai-image-generation-editing-api
AI image generation, editing, and background removal API via Bria.ai — authenticates via OAuth device flow and caches credentials in /.bria/credentials, then calls 20+ endpoints to remove backgrounds for transparent PNGs and cutouts, generate images from text prompts, and edit…
gemini-image-generator
Use when generating professional posed product images for e-commerce using Gemini AI with optimized prompts.
beat-sync-reel
Generates Instagram Reels where product image cuts are synced to audio beats. Accepts audio as a local file, URL, or search query. Uses librosa for beat detection, FFmpeg Ken Burns for scene animation, and Pillow for text overlays. No AI video generation — fully free, fast, and scalable.
ecom-details-image
A planning tool for e-commerce product visuals, including main-image concepts, scene ideas, detail-page directions, and prompts for image-generation systems.
byted-livesaas-master
A control tool for managing business livestreams, including rooms, comments, viewers, product cards, coupons, and live-session settings.
byted-ind-ecom-product-video-prompt
A structured prompt-writing guide for creating e-commerce product videos with Seedance 2.0. It turns one or more product images into a product showcase script using scene settings, timed shots, and output constraints.