ai-podcast-creation

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

• Third-party content exposure detected (low risk: 0.10). The NotebookLM-style workflow feeds a user-supplied <your-document-content> into an LLM prompt, and that document may contain outsider-authored web or third-party text that the agent did not select from a monitored feed, so this is low exposure.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (medium risk: 0.30). The decisive boundary is setup-time installation of skill material from a GitHub locator (npx skills add belt-sh/cli, a blockquote instruction to install the belt CLI skill), and the belt-sh publisher is not independently established or authenticated by digest or signature, so it falls to the common/reputable unofficial tier.