agent-right-brain

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

• Third-party content exposure detected (high risk: 0.80). The skill's references and workflow explicitly accept and fetch arbitrary public URLs for media (e.g., references/stt/dashscope.md shows "rawgenai dashscope stt create https://example.com/meeting.wav" and references/voice/kling.md shows "rawgenai kling voice create ... --audio 'https://example.com/audio.mp3'"), so the agent ingests untrusted third‑party audio/video content which it reads and acts on as part of its workflow.