castmagic

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

• Third-party content exposure detected (high risk: 0.80). The SKILL.md explicitly instructs the agent to use Membrane actions (e.g., "get-transcript" and "create-transcript" via membrane action run) to fetch transcripts and to submit arbitrary audio/video URLs to Castmagic, meaning the agent will ingest untrusted, user-generated third-party content from Castmagic that could influence subsequent actions.