Downloads product images from Google Sheets URLs, resizes with PIL, removes backgrounds, and saves to originals/resized/nobg folders.
npx claudepluginhub alpacalabsllc/skills-for-architects --plugin 06-materials-researchThis skill is limited to using the following tools:
Download product images from a Google Sheet, normalize sizing, and remove backgrounds. Saves output at each processing stage.
Removes backgrounds from single or batch images using AI (rembg/U2-Net) or built-in white-to-transparent methods. Outputs PNG/WebP with transparency for photos, products, icons.
Batch-resizes images for web (WebP 1920/1200/400px), social (center-cropped WebP for Instagram/Twitter/LinkedIn), slides (JPEG 1024x768/1920x1080), print (300 DPI ARCH JPEGs). Outputs to subfolders.
Removes backgrounds from images using each::sense AI API. Creates transparent PNGs, replaces with solid colors or scenes, handles complex edges like hair precisely. For product photos and portraits.
Share bugs, ideas, or general feedback.
Download product images from a Google Sheet, normalize sizing, and remove backgrounds. Saves output at each processing stage.
Works with the master Google Sheet — the 33-column schema defined in ../../schema/product-schema.md. Image URLs are in column AC, product names in column E. Read ../../schema/sheet-conventions.md for CRUD patterns with MCP tools.
If no arguments provided, ask the user:
docs.google.com/spreadsheets/d/{ID}/...). 2. Image URL column — which column contains image URLs (default: AC in the master schema, or the user can specify)E in the master schema). If not provided, derive names from the image URL/filename../product-images-YYYY-MM-DD/ as default but let the user pick any path.Use mcp__google-sheets__list_sheets to inspect the sheet, then mcp__google-sheets__get_sheet_data to read the image URL column and optional name column.
Build a list of { index, url, name } entries. Skip empty rows.
Create the output directory at the user's chosen path with 3 subfolders:
<output-path>/
├── originals/ # Raw downloads
├── resized/ # Normalized sizing
└── nobg/ # Background removed
If the folder already exists, append a suffix: -2, -3, etc.
Download each image using curl in Bash:
curl -L -o "<output-path>" "<url>"
IMPORTANT: Use curl, NOT WebFetch. WebFetch processes content through an AI model which corrupts binary image data.
Name files as: 001-product-name.png, 002-product-name.png, etc.
001-image.png, 002-image.png, etc.If the downloaded file is not a PNG (check extension or content type), convert it to PNG during the resize step.
Run a Python script to resize all images in originals/ → resized/:
from PIL import Image
import os, sys
input_dir = sys.argv[1] # originals/
output_dir = sys.argv[2] # resized/
max_edge = int(sys.argv[3]) if len(sys.argv) > 3 else 2000
for fname in sorted(os.listdir(input_dir)):
if not fname.lower().endswith(('.png', '.jpg', '.jpeg', '.webp', '.gif', '.bmp', '.tiff')):
continue
try:
img = Image.open(os.path.join(input_dir, fname))
img = img.convert("RGBA")
w, h = img.size
longest = max(w, h)
if longest > max_edge:
scale = max_edge / longest
new_w, new_h = int(w * scale), int(h * scale)
img = img.resize((new_w, new_h), Image.LANCZOS)
out_name = os.path.splitext(fname)[0] + ".png"
img.save(os.path.join(output_dir, out_name), "PNG")
print(f"OK: {fname} → {out_name} ({img.size[0]}x{img.size[1]})")
except Exception as e:
print(f"FAIL: {fname} — {e}")
Rules:
Check if rembg is installed. If not, install it:
pip3 install rembg onnxruntime
Then run background removal on all resized images → nobg/:
from rembg import remove
from PIL import Image
import os, sys, io
input_dir = sys.argv[1] # resized/
output_dir = sys.argv[2] # nobg/
for fname in sorted(os.listdir(input_dir)):
if not fname.lower().endswith('.png'):
continue
try:
input_path = os.path.join(input_dir, fname)
with open(input_path, 'rb') as f:
input_data = f.read()
output_data = remove(input_data)
img = Image.open(io.BytesIO(output_data))
img.save(os.path.join(output_dir, fname), "PNG")
print(f"OK: {fname}")
except Exception as e:
print(f"FAIL: {fname} — {e}")
Note: The first run of rembg downloads the u2net model (~170MB). Warn the user this may take a minute.
After processing, print a summary:
## Product Image Processing Complete
📁 Output: ./product-images-YYYY-MM-DD/
| Stage | Success | Failed |
|-------------|---------|--------|
| Downloaded | 12 | 1 |
| Resized | 12 | 0 |
| BG Removed | 12 | 0 |
### Failures
- 003-chair-arm.png: Download failed (404 Not Found)
Include the full path to the output folder so the user can open it.