The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses subprocess.run within scripts/run.py and scripts/setup_environment.py to create a virtual environment, install packages via pip, and execute its internal scripts.
[EXTERNAL_DOWNLOADS]: Initial setup automatically downloads and installs Python packages (patchright, python-dotenv) and the Google Chrome browser to support automation and stealth features.
[INDIRECT_PROMPT_INJECTION]: The skill ingests data from external NotebookLM notebooks, which may contain attacker-controlled content. Evidence chain: (1) Ingestion point: scripts/ask_question.py reads answers from the web DOM of NotebookLM. (2) Boundary markers: No explicit delimiters or safety instructions are wrapped around the retrieved content. (3) Capability inventory: The skill has network access for browsing and file system access for managing notebook metadata and authentication state. (4) Sanitization: The retrieved text is not filtered or sanitized before being presented to the agent.
[DYNAMIC_EXECUTION]: The run.py script serves as a wrapper to dynamically execute other scripts in the skill's directory (auth_manager.py, notebook_manager.py, ask_question.py) ensuring they operate within the isolated Python environment.

notebooklm