2d-pixel-asset

CRITICAL E006: Malicious code pattern detected in skill scripts.

• Malicious code pattern detected (high risk: 1.00). The package contains a clear supply-chain / remote code execution risk: the post-processing script unconditionally loads HuggingFace model code with trust_remote_code=True (allowing arbitrary remote code execution), and the automation instructions direct uploading local files via browser MCP tools (which can exfiltrate images or data if misused).

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 1.00). The skill's process_asset.py can call remove_bg_ml which uses transformers.AutoModelForImageSegmentation.from_pretrained("ZhengPeng7/BiRefNet", trust_remote_code=True) and will download and execute code from the HuggingFace model repository at https://huggingface.co/ZhengPeng7/BiRefNet at runtime, which can run remote code.