The Agent Skills Directory

[COMMAND_EXECUTION]: The library includes logic to execute a user-defined shell command for refreshing authentication cookies. The command is retrieved from the NOTEBOOKLM_REFRESH_CMD environment variable and executed via subprocess.run with shlex.split. This is an intended, opt-in feature for advanced automation workflows.
[EXTERNAL_DOWNLOADS]: The skill performs network operations to fetch data and download generated artifacts (audio, video, PDFs) from trusted Google-owned domains including notebooklm.google.com, storage.googleapis.com, and googleusercontent.com.
[DATA_EXFILTRATION]: The skill manages highly sensitive Google session cookies required for authentication. These are stored locally in ~/.notebooklm/profiles/<profile>/storage_state.json. The implementation enforces restrictive file permissions (0o600, read/write by owner only) to protect these credentials.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from external URLs and web research results which are then processed by the NotebookLM backend and presented to the AI agent.
Ingestion points: Data enters the agent's context through notebooklm source add and notebooklm source add-research (found in src/notebooklm/_sources.py and src/notebooklm/_research.py).
Boundary markers: The skill does not explicitly use markers or "ignore embedded instructions" warnings when processing external data.
Capability inventory: The library can execute shell commands (via the auth refresh hook), perform network operations, and write files to the local system (src/notebooklm/_core.py, src/notebooklm/_artifacts.py).
Sanitization: Content sanitization is primarily handled by the external Google NotebookLM service.

notebooklm