The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill requires the installation of a third-party Python package notebooklm-py from an unverified repository (teng-lin/notebooklm-py) using uv tool install. This package handles sensitive browser-based authentication flows for Google accounts to access NotebookLM services.\n- [CREDENTIALS_UNSAFE]: The instructions describe the use of high-risk authentication methods, including browser session capture for local logins and the use of a --master-token for headless or CI/CD environments. The management of these long-lived credentials by a third-party tool represents a sensitive data handling risk.\n- [INDIRECT_PROMPT_INJECTION]: The skill possesses a significant attack surface for indirect prompt injection as it is designed to ingest data from untrusted sources.\n
Ingestion points: Data enters the agent's context through URL imports (notebooklm source add --url), deep web research agents (notebooklm research add-web), and Google Drive file discovery (notebooklm research add-drive).\n
Boundary markers: There are no instructions provided to the agent to use delimiters or ignore instructions embedded within these imported sources.\n
Capability inventory: The skill enables the agent to perform file system operations (downloading audio/video/json artifacts), network requests, and configuration changes.\n
Sanitization: There is no mention of filtering or sanitizing the content fetched from external URLs before it is queried or summarized by the AI.\n- [COMMAND_EXECUTION]: The skill uses potentially impactful commands like notebooklm doctor --fix, which performs automated repairs on the local environment, and notebooklm mcp install, which modifies the configuration files of other installed AI agent tools.

notebooklm