The Agent Skills Directory

[PROMPT_INJECTION]: The README.md and SKILL.md files include specific, high-priority instructions directed at the AI agent (e.g., "Do exactly this, nothing more") designed to override the agent's normal reasoning process and force immediate execution of installation scripts.
[REMOTE_CODE_EXECUTION]: The scripts/bootstrap.sh script executes a remote installation script for the uv tool from astral.sh using a curl | sh pipeline. While originating from a well-known service, this pattern executes unverified code in the user's shell environment.
[EXTERNAL_DOWNLOADS]: The installation process involves downloading and installing multiple third-party packages and binaries, including notebooklm-mcp-cli, notebooklm-py, and the Chromium browser via Playwright, expanding the local attack surface.
[COMMAND_EXECUTION]: The skill relies heavily on shell command execution via Bash scripts and Python's subprocess module to manage external CLI tools and environment configurations.
[DATA_EXPOSURE]: The skill manages and references sensitive authentication data, specifically Google account session cookies stored in ~/.notebooklm, which are required for the tool's core functionality but represent a sensitive data management responsibility.

notebooklm