The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses subprocess.run within scripts/run.py and scripts/setup_environment.py to manage a local Python virtual environment and facilitate script wrapping. These operations are restricted to the skill's own directory and are used for legitimate environment management.
[EXTERNAL_DOWNLOADS]: The skill automatically installs necessary Python dependencies from the standard PyPI registry and downloads official Google Chrome binaries for browser automation. These activities originate from well-known services and are part of the documented setup process.
[PROMPT_INJECTION]: As a retrieval-based tool, the skill is subject to indirect prompt injection from data ingested from NotebookLM notebooks.
Ingestion points: Document content is read from notebooklm.google.com via scripts/ask_question.py.
Boundary markers: Absent; the raw response from the notebook is returned to the agent.
Capability inventory: The skill allows shell command execution through the run.py wrapper for internal script management.
Sanitization: Content is passed directly to the agent without filtering, which is typical for this use case.
[SAFE]: Sensitive browser session cookies are stored locally in the data/ directory to handle session persistence issues in Playwright. This data is protected by .gitignore and does not leave the local machine except when sent to Google domains during legitimate automation tasks.

notebooklm