notebooklm

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

• Third-party content exposure detected (high risk: 0.90). The skill programmatically opens and queries user-supplied NotebookLM notebooks on notebooklm.google.com (see scripts/ask_question.py which navigates to a --notebook-url and returns NotebookLM/Gemini responses, and SKILL.md/README which instruct Claude to Smart-Add and act on NotebookLM answers), so untrusted, user-generated third‑party content is ingested and explicitly used to drive follow-up queries and synthesis that can change agent behavior.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill launches a browser at runtime and navigates to NotebookLM notebook URLs (e.g. https://notebooklm.google.com/notebook/...) and reads/injects the remote NotebookLM responses into the agent's outputs/context, so external content fetched at that URL directly controls the agent's answers and is required for the skill to function.