voice-generation

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

• Third-party content exposure detected (high risk: 1.00). The skill's SKILL.md includes a "With Reference Image" flow that accepts arbitrary image_urls (public URLs) and instructs the agent to "Generate a voice narration describing what is happening in this image," so the agent ingests untrusted third-party images which can influence its outputs and behavior.