notebooklm

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

• Third-party content exposure detected (medium risk: 0.30). The runtime workflow ingests outsider-authored free text via the user-provided --question (filled into the NotebookLM query box and then used to drive the browser interaction/LLM response) in askNotebookLM(question, notebookUrl) and ask command path.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill directly opens and scrapes NotebookLM at runtime (https://notebooklm.google.com) to obtain answers that control agent output, and the README also instructs executing remote install code from https://bun.sh/install via curl | bash.