diff --git a/skills/archived/atomk-temu-batch-listing/SKILL.md b/skills/archived/atomk-temu-batch-listing/SKILL.md
new file mode 100644
index 0000000..b48aded
--- /dev/null
+++ b/skills/archived/atomk-temu-batch-listing/SKILL.md
@@ -0,0 +1,79 @@
+---
+name: atomk-temu-batch-listing
+category: cross-border-ecommerce
+description: Batch claim products from AtomK (家居百货 category), generate Temu listing data via AI, resize images to 800x800, and export as CSV.
+---
+
+# AtomK → Temu Batch Listing Generator
+
+## Overview
+Claims products from AtomK (九个网) in the 家居百货 (Home & Garden) category, generates Temu-compatible listing data via AI, resizes all images to 800×800 pixels, and exports a complete CSV file.
+
+## Prerequisites
+- AtomK account: `admincao / Tt123456!` (login at `https://www.atomlisting.com/`)
+- Node.js Playwright installed at `/home/ubuntu/.hermes/hermes-agent/node_modules/playwright`
+- Python with Pillow for image resizing
+- DrissionPage is NOT installed; use Playwright only
+
+## Workflow
+
+### 1. Login to AtomK
+```javascript
+// Use Playwright with ignoreHTTPSErrors=true, --no-sandbox
+await page.goto('https://www.atomlisting.com/login');
+await page.fill('input[name="account"]', 'admincao');
+await page.fill('input[name="password"]', 'Tt123456!');
+await page.click('button:has-text("登 录")');
+await page.waitForURL('**/dashboard');
+```
+
+### 2. Claim Products from 家居百货 Category
+- Navigate to product pool page
+- Use `<select>` element to choose "家居百货" category
+- Click claim button for each product
+- Wait for page refresh/navigation between claims
+- Target: 20 products per batch
+- Extract product code (e.g., DEU9E6ZL, AYTRAHXM) from each claimed product
+
+### 3. Generate Temu Listing Data
+For each claimed product:
+- Navigate to product editor page (`/editor/{id}`)
+- Click "AI 生成上架信息" button
+- Select "Temu" platform tab
+- Wait for AI generation to complete (~1-2 min per product)
+- Extract: Product name, Category path, Price, Quantity, Description, Keywords, Image URLs
+- Note: Some products may have fewer images (1-14 images typical)
+
+### 4. Resize Images to 800×800
+- Download all product images from COS URLs (腾讯云 COS: `9websclub-1251422183.cos.ap-hongkong.myqcloud.com`)
+- Resize each image to exactly 800×800 pixels (use white background padding if needed)
+- Save to `/tmp/temu_batch_images_800x800/{product_code}/image_{N}.jpg`
+- Typical batch: 20 products × ~12 images = ~240 images
+
+### 5. Export CSV
+Generate Temu-standard CSV with columns:
+```
+Product name,Category path,Price,Quantity,Product description,Package weight,Package length,Package width,Package height,Keywords,Image URLs,SKU,Product code,UPC,GTIN,EAN
+```
+- Image URLs: semicolon-separated list of COS URLs (or local paths if preferred)
+- Save to `/tmp/temu_batch_listing_final.csv`
+
+### 6. Delivery
+- User's computer has LAN share: `\\192.168.9.12\Public\产品组\csv\`
+- Agent cannot access LAN shares directly; provide CSV content for copy-paste or package as tar.gz for download
+- Package: `tar czf /tmp/temu_batch_20.tar.gz /tmp/temu_batch_listing_final.csv /tmp/temu_batch_images_800x800/`
+
+## Pitfalls
+- **Page refresh timing**: After claiming a product, the page may refresh or redirect. Wait for stable state before next claim.
+- **AI generation timeout**: Each product takes 1-2 min. For 20 products, allow 20-40 min total. Use batch/parallel processing if possible.
+- **Image count varies**: Products have 1-14 images. Don't assume fixed count.
+- **Select dropdown**: Category filter uses native `<select>` element, not custom dropdown. Use `selectOption()` not click.
+- **DrissionPage not installed**: Always use Playwright for this workflow.
+- **Domain**: Use `https://www.atomlisting.com/` NOT `bt109atomk.sh3.ikuai7.com` (latter silently fails login).
+
+## Verification
+- Confirm 20 products claimed successfully
+- Verify all 20 products have Temu data generated
+- Check all images are exactly 800×800 pixels
+- Validate CSV has 20 data rows + 1 header row
+- Ensure all required Temu fields are populated
\ No newline at end of file