notebooklm-hardened
Pass
Audited by Gen Agent Trust Hub on Apr 21, 2026
Risk Level: SAFECOMMAND_EXECUTION
Full Analysis
- [COMMAND_EXECUTION]: The file
scripts/notebooklm.mjsexecutes thenotebooklmsystem command using Node.jsspawnSync. The implementation passes arguments as an array, which is a secure practice that avoids shell interpolation and mitigates potential command injection risks while maintaining intended functionality. - [DATA_EXFILTRATION]: The skill documentation includes comprehensive defensive instructions (guardrails) that explicitly forbid the agent from exfiltrating notebook content or credentials. These instructions provide a clear security policy for the agent to follow when handling sensitive research data locally.
- [PROMPT_INJECTION]: As the skill ingests data from external sources and research queries via the NotebookLM CLI, it possesses an attack surface for indirect prompt injection. However, this risk is addressed by explicit boundary markers and instructions in the
SKILL.mdfile that mandate confirmation for sensitive actions and prohibit the exposure of internal data structures.
Audit Metadata