The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses subprocess.run in scripts/run.py, scripts/setup_environment.py, and scripts/__init__.py to create and manage a local Python virtual environment, install required packages, and execute internal scripts. These operations are limited to the skill's own directory and are necessary for its operation.
[EXTERNAL_DOWNLOADS]: The skill utilizes the patchright library to download and install the Google Chrome browser during its setup phase. This is a functional requirement to enable automated interaction with the NotebookLM web interface.
[PROMPT_INJECTION]: The skill processes responses from Google NotebookLM which represents an indirect prompt injection surface, as the notebook contents are derived from user-provided documents.
Ingestion points: Answer text is extracted from the browser session in scripts/ask_question.py.
Boundary markers: The skill appends a FOLLOW_UP_REMINDER string to queries but does not implement strict delimiters for the external content.
Capability inventory: The skill suite possesses shell execution capabilities via its internal wrapper scripts and standard file system access.
Sanitization: The retrieved notebook content is not sanitized or filtered before being returned to the agent context.

blog-notebooklm