liufelix2004/shoppe-skill/shopee-auto-lister

Shopee Auto Lister Skill

Fully automated pipeline to source phone case products from Pinduoduo competitor stores and list them on Shopee Malaysia — zero manual intervention after setup.

Pipeline Overview

拼多多 competitor store
  ↓  [pdd_scraper.py]      Playwright → Bit Browser CDP → products.json
  ↓  [image_processor.py]  Download + strip watermark + compress → ./images/
  ↓  [ai_processor.py]     DeepSeek API → EN/BM title + description + MYR price → processed.json
  ↓  [shopee_uploader.py]  Playwright → Bit Browser CDP → Shopee seller backend → uploaded.json
  ↓  [scheduler.py]        Windows Task Scheduler → runs daily at 10:00 MYT

Project File Structure

shopee-auto-lister/
├── config.py
├── pdd_scraper.py
├── image_processor.py
├── ai_processor.py
├── shopee_uploader.py
├── scheduler.py
├── products.json
├── processed.json
├── uploaded.json
├── failed.json
└── images/{goods_id}/1.jpg ... 9.jpg

Key Design Decisions

• Bit Browser CDP — never open a new Chrome instance; always connect via

playwright.chromium.connect_over_cdp(f"http://localhost:{port}") to reuse existing proxy/fingerprint and minimise account ban risk

• Random delays — scraper: 2–5s between detail pages; uploader: 15–30s between listings
• Daily cap — max 20 listings/day during store warm-up period
• Stock always 999 — dropshipping model
• Pricing formula — MYR price = CNY cost × 2.8, rounded to x.90

Module Reference

Read the relevant reference file before writing or debugging any module:

Quick Start for Agent

1. Read references/config.md first
2. Build in order: config → scraper → image → ai → uploader → scheduler
3. Verify each module's output file before proceeding to next
4. Never remove random delay logic — it is risk control, not optional
5. When Shopee selectors break, inspect live page via Bit Browser and update references/uploader.md

Common Issues