菜单
Analyze TikTok promotional videos for a known productId. Triggers: product videos, influencer performance, video sales, video GMV, views/likes/shares. Use after a product is known, not for product discovery or market rankings.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
| name | ecommerce.tiktok-product-video |
| version | 1.0.0 |
| category | product-sourcing |
| description | Analyze TikTok promotional videos for a known productId. Triggers: product videos, influencer performance, video sales, video GMV, views/likes/shares. Use after a product is known, not for product discovery or market rankings. |
Before running this skill, if the user provides a product, image, ASIN, listing, product ID, or product idea, first identify:
Use this product understanding to choose parameters, filters, competitors, regions, keywords, or analysis dimensions before executing the script.
Which promotional videos and influencers are driving views, engagement, sales, or GMV for a specific TikTok Shop product?
productId and asks for product videos, influencer performance, video sales, video GMV, views, likes, comments, shares, or publish timing.If the user request is broad or ambiguous, identify:
productId; if missing, first use product search or top-selling ranking to find oneUse sensible defaults when the user's intent is clear: sort by views descending, pageSize 50, no creator filter.
ecommerce.tiktok-product-research when the user needs to search for TikTok products or find a productId.ecommerce.tiktok-top-selling when the user asks for market/category best-seller rankings.productId.Default output should include:
This skill queries promotional videos associated with a TikTok Shop product, helping sellers analyze video marketing performance and identify effective influencer content strategies.
This tool retrieves the list of promotional videos linked to a specific TikTok product. Each video record includes engagement metrics (views, likes, comments, shares, favorites), estimated sales attribution (video sales count and GMV), video metadata (duration, resolution, publish date), and the creator (influencer) ID. This enables sellers to understand which videos drive the most sales for a product and what content patterns work best.
Required input: A productId is mandatory. You can obtain product IDs from other TikTok product search or ranking tools.
Sort fields: Videos can be sorted by views (1), likes (2), shares (3), video sales (4), video GMV (5), or publish date (6).
Pagination: pageSize must be a multiple of 10, max 100. The backend fetches in batches of 10 and merges results.
| Parameter | Type | Required | Description | Default |
|---|---|---|---|---|
| productId | string | Yes | TikTok product ID | - |
| userId | string | No | Filter by influencer ID | - |
| productVideoSortField | integer | No | Sort field: 1=views, 2=likes, 3=shares, 4=video sales, 5=video GMV, 6=publish date | 1 |
| sortType | integer | No | Sort order: 0=ascending, 1=descending | 1 |
| minCreateTime | integer | No | Video publish start time (Unix timestamp in seconds) | - |
| maxCreateTime | integer | No | Video publish end time (Unix timestamp in seconds) | - |
| pageNum | integer | No | Page number | 1 |
| pageSize | integer | No | Results per page (multiple of 10, max 100) | 50 |
This tool calls the NexScope proxy API. See references/api.md for calling conventions, request parameters, and response structure. You can also execute scripts/tiktok_product_video.py directly to run queries.
1. Top videos by views for a product
{
"productId": "1729382310407603945",
"productVideoSortField": 1,
"sortType": 1,
"pageSize": 20
}
2. Find highest-converting videos (by video sales)
{
"productId": "1729382310407603945",
"productVideoSortField": 4,
"sortType": 1,
"pageSize": 20
}
3. Videos by a specific influencer for a product
{
"productId": "1729382310407603945",
"userId": "7234567890123456789",
"productVideoSortField": 1,
"sortType": 1
}
4. Recent videos in a time range (sorted by GMV)
{
"productId": "1729382310407603945",
"minCreateTime": 1717200000,
"maxCreateTime": 1719792000,
"productVideoSortField": 5,
"sortType": 1
}
5. Videos sorted by publish date (newest first)
{
"productId": "1729382310407603945",
"productVideoSortField": 6,
"sortType": 1,
"pageSize": 50
}
officialUrl is available, provide it so users can view the video on TikTokcoverUrl is present, mention it so the user knows video thumbnails are availableduration (seconds) to a readable format (e.g., "1:30" for 90 seconds)hashTag when present to help users understand content themesplayAddr field may expire quickly; prefer officialUrl for sharing| User Says | Scenario |
|---|---|
| "Show me the promotional videos for this TikTok product" | Query videos by product ID |
| "Which videos are driving the most sales for this product" | Sort by video sales (field 4) |
| "What influencer videos are promoting this product" | General video list query |
| "Show me videos by a specific creator for this product" | Filter by userId |
| "Recent promotional videos for this product" | Filter by time range or sort by date |
| "Which TikTok videos have the highest GMV for this product" | Sort by video GMV (field 5) |
| "Analyze the video marketing performance of this product" | Comprehensive video list query |
When users ask about "TikTok videos", determine whether they want videos associated with a specific product (this skill) or general TikTok video analytics (not this skill). If they mention a product ID or ask "what videos are promoting product X", this skill applies. If they ask about trending videos in general without a product context, this skill does not apply.
To avoid overflowing the agent context, persist the response to disk and extract only the fields you need:
python scripts/response_io.py run --script scripts/tiktok_product_video.py --out-dir <DIR> '<params>'
python scripts/response_io.py read <file> --fields "<paths>" # or --path "<JMESPath>"
Pick
--out-diroutside any git working tree (e.g./tmp/...on Unix,%TEMP%/...on Windows). Persisted responses may contain PII, pricing, or auth-sensitive data -- do not commit them. Files are not auto-deleted; clean up when the task is done.
run writes the full response to a file and emits only a schema preview + file path. read projects specific fields, with --limit/--offset for slicing and --format json|jsonl|csv|table for output.
When to prefer this pattern -- apply your judgment based on the response characteristics, e.g.:
For small, single-use responses, calling the main script directly is fine.
The preview is a truncated schema + sample, not the full data. Any field-level decision must read from the persisted file via read.
| Variable | Required | Description |
|---|---|---|
NEXSCOPE_PROXY_BASE | Yes | NexScope proxy base URL |
NEXSCOPE_API_KEY | Yes | NexScope proxy API key |
Compare Amazon competitor business metrics: price, reviews, BSR, sales, seller count, traffic. Triggers: competitor metrics, who performs best, price/BSR/review comparison. Use for business performance, not listing copy; use competitor-listing-analyzer for content structure.
Analyze Amazon listing content and structure: title, bullets, images, A+ content, and listing quality. Triggers: listing breakdown, reverse-engineer listings, title/bullet/image analysis. Use for content audits, not business metrics; use competitor-analyzer for price/BSR/sales.
Validate whether demand is real by comparing search volume with sales and buying intent. Triggers: is demand real, fake demand, inflated volume, are people buying, conversion gap. Use for demand proof, not broad market overview or niche go/no-go scoring.
Check visual/design patent similarity from product images. Triggers: design infringement, look-alike patent, visual similarity, design clearance. Use for appearance/design risk, not utility claims or legal-status checks.
Suggest product differentiation and USP using competitor/review evidence. Triggers: how to stand out, USP, competitive advantage, product improvement. Use for positioning and feature strategy, not raw competitor metrics or review monitoring.
Find visually similar Amazon products from an image. Triggers: reverse image search, find lookalikes, who else sells this, source by image. Use for marketplace lookalikes, not patent similarity or design infringement checks.