The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes a run.py wrapper and setup_environment.py script to automatically create a Python virtual environment and install dependencies using subprocess.run.
[COMMAND_EXECUTION]: The initialization process in scripts/__init__.py and scripts/setup_environment.py executes system commands to install the patchright automation package and Google Chrome browser binaries.
[EXTERNAL_DOWNLOADS]: The skill downloads and installs external software, specifically the Google Chrome browser and the patchright library, during the environment setup phase to facilitate browser automation.
[PROMPT_INJECTION]: The skill includes explicit instructions in SKILL.md (Follow-Up Mechanism) that direct the agent to analyze output for specific markers and potentially initiate additional queries, which guides the agent's decision-making process based on external script output.
[PROMPT_INJECTION]: As the skill retrieves and displays content from external Google NotebookLM documents, it is vulnerable to indirect prompt injection if those documents contain instructions designed to manipulate the agent's behavior during processing.
[DATA_EXFILTRATION]: The skill stores sensitive browser session data, including cookies and storage state, in the local directory ~/.claude/skills/notebooklm/data/browser_state/ to maintain authentication across sessions.

notebooklm