face-swap

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (high risk: 0.80). The SKILL.md clearly requires passing arbitrary image URLs via the top-level "image_urls" parameter (see the Quick Start/curl and Python examples), so the agent fetches and processes external, user-supplied images from the open web which are untrusted and can materially influence its decisions about which faces/targets to modify.