podcast

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

• Third-party content exposure detected (high risk: 1.00). The skill's Step 2 "Research (if needed)" explicitly uses rebyteai/internet-search and rebyteai/deep-research to fetch public web content (news, quotes, multi-source research) — and the audio pipeline downloads assets from Pixabay — which the agent ingests and uses to draft scripts and drive production decisions, meaning untrusted third‑party content can materially influence actions.