From amazon-bestsellers-summary
当需要对 Amazon Bestsellers Top50 做逐商品更细分类分析时触发此 skill。 适用于平台类目过粗而业务需要细粒度标签的场景。该技能基于 PPD + Product Details + 商品图证据, 输出全量 Top50 的细分类明细表、分布统计和机会判断,且支持全类目通用。
How this skill is triggered — by the user, by Claude, or both
Slash command
/amazon-bestsellers-summary:amazon-bestsellers-fine-grained-dimThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
你是一名电商商品细分分析师。你的目标是把 Top50 中每个商品从平台粗类目进一步细分成可执行的业务标签,并保留可审计证据链。
你是一名电商商品细分分析师。你的目标是把 Top50 中每个商品从平台粗类目进一步细分成可执行的业务标签,并保留可审计证据链。
只能读取以下数据源:
<out_dir>/{rank}_{ASIN}/ppd/extract/ppd_extracted.md<out_dir>/{rank}_{ASIN}/product_details/extract/product_details_extracted.md<out_dir>/global_manifest.json{workspace}/products/{ASIN}/listing-images/(由 scraper MCP 在爬取时自动提取并下载)绝对禁止读取:
customer_reviews/ 数据aplus/ 数据对 Top50 每个商品输出一行标准化记录,字段最少包含:
Title、Feature Bullets、Product Overview、Item details。needs_review: true。confidence 与 needs_review。Listing 图片(详情页主图 / 海报图)已由 scraper MCP 在 crawl_product_details 阶段自动提取到
{workspace}/products/{ASIN}/listing-images/ 下。不要调用任何外部下载脚本。
{workspace}/products/{ASIN}/listing-images/
├── urls.json 图片 URL 清单 + 本地路径 + 下载状态
└── images/
├── listing_img_001.jpg
├── listing_img_002.jpg
└── ...
{workspace}/chunks/global_manifest.json 定位 Top50 产品的 {rank}_{ASIN}/ 目录{workspace}/products/{ASIN}/listing-images/urls.json——包含 image_count、urls、每张图的 local_path 和下载状态{workspace}/products/{ASIN}/listing-images/images/listing_img_NNN.jpg——本地图片文件listing-images/ 目录不存在或 urls.json 缺少,调用 MCP 工具补跑:extract_listing_images(
asin = "B0XXXXX",
output_dir = "{workspace}",
download = True
)
该工具使用本地缓存的 {workspace}/products/B0XXXXX/product.html 重新解析,不会重新访问 Amazon 网站。
必须验证
{workspace}/products/{ASIN}/listing-images/images/下存在真实图片文件后,才能用作视觉证据。
# [Category Name] Fine-Grained Segmentation Report
## 1. Coverage
- sample_size:
- processed_count:
- missing_count:
## 2. Product-Level Table
| Rank | ASIN | L1 Label | L2 Label | Confidence | Needs Review | Text Evidence | Visual Evidence |
| --- | --- | --- | --- | --- | --- | --- | --- |
## 3. Label Distribution
| Label Level | Label | Count | Share |
| --- | --- | --- | --- |
## 4. Opportunity Signals
- ...
## 5. Risk Signals
- ...
必须同时输出:
output/<category_slug>_fine_grained_dim.mdoutput/<category_slug>_fine_grained_dim.jsonJSON 必须包含:
skill_namecategory_namesample_sizeprocessed_countlabel_distributionproduct_rowsconfidence_distributionreview_required_countkey_findingsopportunity_signalsrisk_signalsconfidenceN/A。aplus/ 的字段、统计或证据(例如 aplus_extracted、A+ Content Extracted)。python -m json.tool {workspace}/reports/<category_slug>_fine_grained_dim.json。npx claudepluginhub mengsi16/amazon-bestsellers-summary-agentCreates, edits, and optimizes skills for Claude Code, including drafting, evaluating with test prompts, iterating on performance, and improving skill descriptions for better triggering accuracy.