Security audit
Security checks for vulnerabilities and agentic risk
This skill is a straightforward image object-recognition prompt with disclosed visual analysis behavior and no executable code or persistence.
Install only if you are comfortable sending image URLs for visual analysis. Avoid using it on images containing documents, credentials, private rooms, valuables, medications, or other sensitive details unless the people affected have consented, and prefer cropping images to the objects you actually want identified.
The skill sends or processes user-supplied image content for detailed object recognition, including inferred brand, material, and price, but provides no privacy notice about possible exposure of sensitive visual data. Users may unknowingly submit images containing personal belongings, documents, location clues, or other confidential details, creating avoidable privacy and data-handling risk.
The 'professional' mode explicitly expands analysis to small objects and drawer interiors, which increases the likelihood of extracting highly sensitive household or workplace information from images. Without guardrails or warnings, this feature can facilitate overcollection of private details such as valuables, credentials, medications, or business materials beyond normal user expectations.
The natural-language instructions and descriptive sections are written in Chinese, which may impose a language constraint on users without any opt-in or explanation. The policy requires either offering language choice or documenting a justified locale-specific limitation, neither of which is present here.
