Install any skill in seconds. Free to start, no credit card required.
Get Started Free →跨境电商全链路自动化工具。集成1688采集、智能清洗、多平台上架(微信小店/Shopify/TikTok)、推广方案(关键词/竞品分析/广告文案)、短视频创作(MoviePy竖屏视频)、一键代发、爆品挖掘(趋势聚合+6维评分)、闲鱼二手选品捡漏(品牌识别/虚标过滤/捡漏评分/价格监控)、全自动流水线(挖掘→采集→清洗→上架→推广→视频)。
.claude/skills/anbeime-ecommerce-full-pipeline/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | -11% | 0% |
| case-02 | ✗→✓ | ▲ Improved | 43% | 0% |
| case-03 | ✗→✓ | ▲ Improved | 0% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 24% | 0% |
| case-08 | ✗→✓ | ▲ Improved | 45% | 0% |
bashpip install requests beautifulsoup4 lxml pyyaml pytrends moviepy pillow
bash# 自动挖掘爆品 → 采集 → 清洗 → 上架 → 推广 → 视频生成 python main.py auto-pipeline --seed-keywords "收纳盒,蓝牙耳机" --top-n 5 --product-limit 3 --platform all
bash# Step 1: 采集 python main.py scrape --keyword "蓝牙耳机" --pages 2 --detail --download-images # Step 2: 清洗(过滤毛利率低于20%、供应商评分低于60分的商品) python main.py clean --input data/raw_products.json --min-margin 0.20 --min-score 60 # Step 3: 上架(生成多平台模板) python main.py publish --input data/cleaned_products.json --platform all --export all # Step 4: 推广(关键词+竞品分析+广告文案+预算) python main.py promote --input data/cleaned_products.json --limit 10 --daily-budget 50 # Step 5: 生成推广短视频 python main.py video --input data/cleaned_products.json --limit 5 --duration 15
bash# 自动挖掘趋势爆品 python main.py trend --categories electronics,home --top-n 10 # 指定种子词挖掘 python main.py trend --seed-keywords "充电宝,手机壳" --top-n 10
bash# 闲鱼搜索 python main.py xianyu --keyword "投影仪" --price-max 999 --pages 3 --condition 9成新 # 捡漏搜索(带参数需求) python main.py hunt --keyword "投影仪 4K 云台" --budget 999 \ --min-lumens 1000 --require-4k --require-gimbal --require-wall # 智能推荐(多关键词综合) python main.py recommend --keywords "投影仪,家用投影,4K投影" --budget 999 \ --min-lumens 1000 --require-4k --require-gimbal # 价格监控 python main.py monitor --keywords "投影仪,坚果投影" --budget 999
bashpython main.py fulfill --source-url "https://detail.1688.com/offer/xxx.html" \ --receiver-name "张三" --receiver-phone "13800138000" \ --receiver-address "北京市朝阳区xxx" --sku-spec "白色/大号" --quantity 1
爆品评分 = 趋势速度 × 35% + (1-竞争度) × 25% + 利润潜力 × 25% + 供应稳定 × 15%
| 维度 | 权重 | 评分逻辑 | |------|------|---------| | 趋势速度 | 35% | Google Trends 搜索量变化率,近期涨幅越高分越高 | | 竞争度 | 25% | 市场竞争强度反向指标,竞争越低分越高 | | 利润潜力 | 25% | 1688 供货价与预估售价的利润率 | | 供应稳定 | 15% | 供应商数量和供货稳定性 |
评分公式:价格优势(30分) + 性价比(20分) + 成色(15分) + 卖家信誉(15分) + 参数匹配(20分) - 风险扣分
| 维度 | 分值 | 评分逻辑 | |------|------|---------| | 价格优势 | 0-30 | 实际价/合理二手价的比值越低分越高,降价幅度加分 | | 性价比 | 0-20 | 相对全新价的折扣率,折扣越大分越高 | | 成色 | 0-15 | 全新15分 > 几乎全新13 > 99新12 > 9成新10 > 正常使用7 | | 卖家信誉 | 0-15 | 百分百好评+10,想要人数加分 | | 参数匹配 | 0-20 | 按用户需求参数(亮度/4K/云台等)的匹配度 | | 风险扣分 | 0-30 | 低价高参数虚标、无品牌、商家批量出货等风险因素 |
ecommerce-tool/
├── main.py # CLI 入口(13 个子命令)
├── config.yaml # 配置文件
├── requirements.txt # 依赖
├── ecommerce_tool/
│ ├── scraper.py # 1688 采集模块
│ ├── cleaner.py # 数据清洗模块
│ ├── publisher.py # 多平台上架模块
│ ├── promoter.py # 推广方案模块
│ ├── video_generator.py # 短视频生成模块
│ ├── fulfillment.py # 一键代发模块
│ ├── trend_detector.py # 爆品挖掘模块
│ ├── xianyu_scraper.py # 闲鱼采集模块
│ ├── bargain_hunter.py # 捡漏分析模块
│ └── utils.py # 工具函数| 命令 | 说明 | |------|------| | scrape | 从 1688 采集商品 | | clean | 清洗商品数据 | | publish | 生成上架模板(微信小店/Shopify/TikTok) | | promote | 生成推广方案(关键词/竞品/文案/预算) | | video | 生成推广短视频 | | fulfill | 一键代发下单 | | pipeline | 全流程自动化(采集→清洗→上架→推广→视频) | | trend | 自动挖掘爆品趋势 | | auto-pipeline | 全自动爆品流水线(挖掘→采集→上架→推广→视频) | | xianyu | 闲鱼商品搜索 | | hunt | 闲鱼捡漏搜索 | | recommend | 智能推荐(多关键词综合) | | monitor | 价格监控 |
pythonfrom ecommerce_tool.scraper import Scraper1688 from ecommerce_tool.cleaner import DataCleaner from ecommerce_tool.publisher import Publisher from ecommerce_tool.promoter import Promoter from ecommerce_tool.video_generator import VideoGenerator from ecommerce_tool.fulfillment import FulfillmentEngine from ecommerce_tool.trend_detector import TrendDetector from ecommerce_tool.xianyu_scraper import XianyuScraper from ecommerce_tool.bargain_hunter import BargainHunter # 1688 采集 scraper = Scraper1688(config, logger) products = scraper.search("蓝牙耳机", pages=2) # 爆品挖掘 detector = TrendDetector(config, logger) candidates = detector.discover(categories=["electronics"], top_n=10) # 闲鱼捡漏 hunter = BargainHunter(config, logger) results = hunter.hunt( keyword="投影仪", price_max=999, spec_requirements={ "min_lumens": 1000, "support_4k": True, "has_gimbal": True, "support_wall": True, }, ) print(hunter.display_results(results))
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 36,256 | 14,803 | -59% | 1 | 1 | 0% | 6,217 | 5,538 | -11% | 0 | 0 | — |
case-02 | fail→pass | 21,856 | 17,564 | -20% | 1 | 1 | 0% | 4,068 | 5,814 | +43% | 0 | 0 | — |
case-03 | fail→pass | 19,929 | 8,302 | -58% | 1 | 1 | 0% | 4,415 | 4,421 | +0% | 0 | 0 | — |
case-04 | pass→pass | 13,558 | 10,692 | -21% | 1 | 1 | 0% | 2,331 | 4,431 | +90% | 0 | 0 | — |
case-05 | pass→fail | 11,771 | 38,534 | +227% | 1 | 1 | 0% | 2,394 | 6,208 | +159% | 0 | 0 | — |
case-06 | fail→fail | 11,184 | 12,959 | +16% | 1 | 1 | 0% | 1,399 | 4,970 | +255% | 0 | 0 | — |
case-07 | fail→pass | 16,272 | 5,980 | -63% | 1 | 1 | 0% | 2,884 | 3,570 | +24% | 0 | 0 | — |
case-08 | fail→pass | 21,253 | 8,744 | -59% | 1 | 1 | 0% | 2,752 | 3,998 | +45% | 0 | 0 | — |
case-09 | fail→pass | 13,572 | 12,019 | -11% | 1 | 1 | 0% | 2,412 | 4,899 | +103% | 0 | 0 | — |
case-10 | fail→pass | 7,129 | 3,919 | -45% | 1 | 1 | 0% | 1,300 | 3,412 | +162% | 0 | 0 | — |
case-20 | pass→pass | 15,666 | 13,739 | -12% | 1 | 1 | 0% | 2,571 | 4,980 | +94% | 0 | 0 | — |
case-11 | fail→pass | 8,059 | 3,823 | -53% | 1 | 1 | 0% | 1,422 | 3,055 | +115% | 0 | 0 | — |
case-12 | fail→pass | 12,319 | 5,760 | -53% | 1 | 1 | 0% | 2,286 | 3,629 | +59% | 0 | 0 | — |
case-13 | fail→pass | 14,178 | 6,994 | -51% | 1 | 1 | 0% | 2,592 | 3,651 | +41% | 0 | 0 | — |
case-14 | fail→pass | 8,874 | 7,675 | -14% | 1 | 1 | 0% | 1,607 | 4,070 | +153% | 0 | 0 | — |
case-15 | fail→pass | 19,338 | 4,220 | -78% | 1 | 1 | 0% | 2,698 | 3,376 | +25% | 0 | 0 | — |
case-21 | pass→pass | 11,841 | 4,081 | -66% | 1 | 1 | 0% | 1,763 | 3,243 | +84% | 0 | 0 | — |
case-16 | fail→fail | 12,593 | 3,226 | -74% | 1 | 1 | 0% | 1,769 | 3,157 | +78% | 0 | 0 | — |
case-17 | fail→pass | 19,943 | 2,929 | -85% | 1 | 1 | 0% | 1,803 | 3,182 | +76% | 0 | 0 | — |
case-18 | fail→pass | 11,610 | 4,162 | -64% | 1 | 1 | 0% | 2,158 | 3,373 | +56% | 0 | 0 | — |
case-19 | pass→pass | 9,276 | 6,982 | -25% | 1 | 1 | 0% | 1,878 | 3,613 | +92% | 0 | 0 | — |
case-22 | fail→pass | 18,342 | 13,285 | -28% | 1 | 1 | 0% | 2,681 | 4,754 | +77% | 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. 22 cases were attempted. The headline lift of +64 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.