name: image-studio displayName: "Image Studio — 批次影像處理" allowed-tools: Bash(python *) description: > Batch image processing tool supporting remove background (摳圖), replace background (換背景), add watermark (水印), and resize/crop (尺寸裁剪). Processes single images or entire directories in batch mode. Supported formats: PNG, JPG, JPEG, WEBP, BMP, TIFF. Uses a local Python CLI (image_studio.py) backed by rembg + Pillow — no API keys or network required after dependency install. triggers: image-studio, 摳圖, 換背景, 水印, 裁剪, remove background, replace background, watermark, resize image, batch image processing homepage: "" license: MIT

Image Studio — 批次摳圖，換背景，水印新增、尺寸裁剪

A local CLI tool for batch image processing. All operations run offline on your machine using Python — no cloud API, no data leaves your computer.

Quick links: remove-bg · replace-bg · watermark · resize

Skill location

The Python CLI lives at:

~/.agents/skills/image-studio/image_studio.py

All commands use python <path>/image_studio.py <command> ....

First-time setup

# Install the one external dependency
python ~/.agents/skills/image-studio/image_studio.py install

Or manually:

pip install rembg

rembg is only needed for remove-bg and replace-bg. The other commands (watermark, resize, info) work with just Pillow (pre-installed).

Commands

1. 摳圖 — remove-bg

Remove background from image(s). Uses AI-based rembg (u2net model).

# Single image
python ~/.agents/skills/image-studio/image_studio.py remove-bg \
  -i input.jpg \
  -o ./output

# Batch — entire directory
python ~/.agents/skills/image-studio/image_studio.py remove-bg \
  -i ./images/ \
  -o ./output_nobg

# Chinese alias
python ~/.agents/skills/image-studio/image_studio.py 摳圖 \
  -i photo.jpg \
  -o ./output

Fallback: If rembg is not installed, the skill falls back to OpenCV GrabCut (less accurate but dependency-free).

2. 換背景 — replace-bg

Remove background and composite the subject onto a new background (image or solid color).

# Replace with image
python ~/.agents/skills/image-studio/image_studio.py replace-bg \
  -i portrait.jpg \
  -b beach_background.jpg \
  -o ./output

# Replace with solid color
python ~/.agents/skills/image-studio/image_studio.py replace-bg \
  -i product.jpg \
  --bg-color "#3498db" \
  -o ./output

# Batch replace
python ~/.agents/skills/image-studio/image_studio.py replace-bg \
  -i ./photos/ \
  -b studio_bg.jpg \
  -o ./output_newbg

-b and --bg-color are mutually exclusive.

3. 水印 — watermark

Add text or image watermark with configurable position, opacity, and size.

# Text watermark
python ~/.agents/skills/image-studio/image_studio.py watermark \
  -i photo.jpg \
  -t "© 2024 Your Name" \
  -p bottom-right \
  --opacity 0.6 \
  --font-size 48 \
  --font-color "#ffffff" \
  -o ./output

# Image watermark (logo)
python ~/.agents/skills/image-studio/image_studio.py watermark \
  -i photo.jpg \
  --image logo.png \
  --scale 0.15 \
  -p bottom-right \
  --opacity 0.8 \
  -o ./output

# Tiled watermark (full coverage)
python ~/.agents/skills/image-studio/image_studio.py watermark \
  -i document.jpg \
  -t "DRAFT" \
  -p tile \
  --opacity 0.3 \
  -o ./output

# Batch watermark
python ~/.agents/skills/image-studio/image_studio.py watermark \
  -i ./photos/ \
  -t "© 2024" \
  -p bottom-right \
  -o ./output_wm

-t and --image are mutually exclusive.

4. 尺寸裁剪 — resize

Resize or crop image(s) to target dimensions. Four modes:

# Cover mode (default) — best for thumbnails/banners
python ~/.agents/skills/image-studio/image_studio.py resize \
  -i photo.jpg \
  -w 1920 -h 1080 \
  --mode cover \
  -o ./output

# Contain mode — with letterbox bg
python ~/.agents/skills/image-studio/image_studio.py resize \
  -i photo.jpg \
  -w 800 -h 800 \
  --mode contain \
  --bg-color "#000000" \
  -o ./output

# Crop mode — center crop to ratio
python ~/.agents/skills/image-studio/image_studio.py resize \
  -i photo.jpg \
  -w 1080 -h 1080 \
  --mode crop \
  -o ./output

# Batch resize all images in a folder
python ~/.agents/skills/image-studio/image_studio.py resize \
  -i ./images/ \
  -w 800 -h 600 \
  --mode cover \
  -o ./output_resize

5. 資訊 — info

Show image metadata (dimensions, mode, file size).

python ~/.agents/skills/image-studio/image_studio.py info \
  -i photo.jpg

python ~/.agents/skills/image-studio/image_studio.py info \
  -i ./images/

Batch processing patterns

All commands accept directories as -i input. They scan for supported formats: .png, .jpg, .jpeg, .webp, .bmp, .tiff.

Pipeline: batch remove background → replace → resize → watermark

SKILL=~/.agents/skills/image-studio/image_studio.py
DIR=./batch

# Step 1: Remove backgrounds
python $SKILL remove-bg -i $DIR/photos -o $DIR/step1_nobg

# Step 2: Replace with new background
python $SKILL replace-bg -i $DIR/step1_nobg -b bg.jpg -o $DIR/step2_newbg

# Step 3: Resize to uniform size
python $SKILL resize -i $DIR/step2_newbg -w 1920 -h 1080 --mode cover -o $DIR/step3_resize

# Step 4: Add watermark
python $SKILL watermark -i $DIR/step3_resize -t "© 2024" -p bottom-right -o $DIR/final

Error handling

First-use checklist

1. Install rembg (once): python ~/.agents/skills/image-studio/image_studio.py install
2. Verify: python ~/.agents/skills/image-studio/image_studio.py info -i any_image.jpg

   小蔥技能站7w4.net每天更新，海量AI技能等你發現。

How to invoke this skill

The user activates this skill by typing /image-studio or any trigger keyword. When invoked:

1. Check if rembg is available if the task involves background removal. If not, suggest running pip install rembg or the install sub-command.
2. Determine which operation(s) the user needs (remove-bg, replace-bg, watermark, resize).
3. For the requested operation, construct the appropriate python image_studio.py ... command.
4. Run the command and report results to the user (number of images processed, output location).
5. If the user needs a multi-step pipeline, chain commands sequentially.

All output preserves RGBA transparency where applicable. PNG is the default output format for transparency support.

這是一個功能豐富、質量較高的影像處理 Skill，核心功能齊全且支援批次操作，離線執行無需網路。文件清晰易懂，中文命令對國內使用者友好。rembg 缺失時能自動降級保證了可用性。主要不足是缺少測試和示例檔案，且水印功能在某些環境下可能無法正確顯示中文。總體而言是個好用的工具，適合需要批次處理圖片的使用者。