Bg Remove

智能识别并移除图片背景，支持多格式批量处理，输出带透明通道的PNG，可选本地或云端模型。

Security Scan

The skill appears to do what it says (local image background removal) and contains no obvious exfiltration, but there are small inconsistencies (claims about cloud models and extra dependencies) and an implicit model download behavior that you should verify before trusting the skill.

Purpose & Capability

Name/description, script, and metadata consistently describe a local image background-removal tool using rembg/Pillow. However, SKILL.md and descriptions mention support for 'cloud models' and list onnxruntime as a dependency while _meta.json and the script only declare rembg and pillow. The script exposes model choices including 'RMBG-1.4' but contains no explicit logic for selecting a remote/cloud endpoint. This is a minor mismatch between claimed capabilities (cloud option, onnxruntime) and what the package actually implements.

Instruction Scope

SKILL.md and the Python script are focused on reading image files and writing PNG outputs. The runtime instructions and code only reference input/output paths and do not read environment variables, unrelated local files, or system secrets. No instructions direct data to external endpoints. The SKILL.md is fairly specific and does not grant broad discretionary access.

Install Mechanism

There is no install spec (instruction-only with an included script), so nothing is automatically written by a platform installer. SKILL.md and the script indicate that required Python packages must be installed and that '首次使用会自动下载模型文件' (first use will auto-download model files). That implies network activity to fetch model weights at runtime (handled by rembg or related libs), which is expected for model-based tools but is an implicit operation you should be aware of. No install URLs or arbitrary archives are included.

Credentials

The skill requests no environment variables, no credentials, and only file read/write permissions which are required for image processing. The number and type of permissions are proportionate to the stated purpose.

Persistence & Privilege

The skill is not always-enabled and does not request elevated platform persistence. It does not attempt to modify other skills or system-wide configuration. Autonomous invocation is allowed (platform default) but not combined here with other concerning privileges.

What to consider before installing

This skill's code matches its purpose: local image background removal using rembg and Pillow. Before installing or running it, consider: (1) rembg (or its dependencies) may automatically download model weights from the internet on first run — if you need to control network access or trust the model source, run in a sandbox or pre-download vetted model files; (2) SKILL.md mentions onnxruntime and 'cloud models' but the package metadata and script don't implement cloud endpoints or declare onnxruntime as a dependency — if you rely on those features, ask the author for clarification or verify behavior; (3) install dependencies in a virtualenv and inspect where models are downloaded from (rembg documentation) to confirm you are comfortable with the remote hosts. If you are unsure about automatic downloads or running third-party Python packages, test the skill on non-sensitive images in an isolated environment first.

图片智能去背景工具，使用 RMBG-1.4 或 rembg 库实现高精度背景移除。

功能

• 智能去背景: 使用 AI 模型自动识别并移除图片背景
• 支持多种格式: PNG、JPG、WebP、BMP 等常见图片格式
• 批量处理: 支持批量处理多张图片
• 透明背景: 输出带透明通道的 PNG 图片
• 支持本地模型: 可选择使用本地模型或默认模型

触发词

• 去背景、移除背景、图片去背景
• 抠图、去除图片背景、透明背景
• bg-remove、remove background

使用方法

基本用法

# 去除单张图片背景
/bg-remove --input photo.jpg

# 指定输出路径
/bg-remove --input photo.jpg --output result.png

# 批量处理整个文件夹
/bg-remove --input ./photos --output ./output

# 使用特定模型
/bg-remove --input photo.jpg --model RMBG-1.4

高级选项

# 调整处理阈值 (0-1)
/bg-remove --input photo.jpg --threshold 0.5

# 保持原始文件名
/bg-remove --input ./photos --output ./output --keep-name

# 覆盖已存在的文件
/bg-remove --input photo.jpg --force

命令参数

参数	简写	说明	默认值
--input	-i	输入图片或文件夹路径	必填
--output	-o	输出路径（文件或文件夹）	同输入目录
--model	-m	使用的模型 (u2net/RMBG-1.4)	u2net
--threshold	-t	分割阈值 (0-1)	0.5
--keep-name	-k	保持原始文件名	false
--force	-f	覆盖已存在的文件	false
--alpha-matting	-	启用 Alpha Matting 增强边缘	false
--alpha-matting-foreground	-	Alpha Matting 前景阈值	240
--alpha-matting-background	-	Alpha Matting 背景阈值	10

输出说明

• 输出格式为 PNG，支持透明通道
• 文件名默认添加 _nobg 后缀
• 处理后的图片可直接用于设计、电商、产品展示等场景

示例

# 电商产品图去背景
/bg-remove --input ./product.jpg --output ./product_nobg.png

# 批量处理证件照背景
/bg-remove --input ./photos/ --output ./results/

# 人像照片去背景
/bg-remove --input portrait.jpg --model RMBG-1.4 --alpha-matting

依赖

• Python 3.8+
• rembg
• pillow
• onnxruntime

首次使用会自动下载模型文件。