The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses subprocess.run and os.execv in scripts/run.py, scripts/setup_environment.py, and scripts/__init__.py. These calls are used for environment management, such as creating a virtual environment, installing dependencies, and ensuring scripts execute using the correct interpreter.
[EXTERNAL_DOWNLOADS]: The skill downloads the patchright library and Google Chrome/Chromium browsers to perform its stated purpose of browser automation. These downloads originate from standard package registries and official vendor sources.
[INDIRECT_PROMPT_INJECTION]: The skill processes content from an external source (Google NotebookLM). It mitigates potential prompt injection risks by using explicit boundary markers (--- BEGIN UNTRUSTED NOTEBOOKLM CONTENT ---) and persisting the untrusted data to private local files (0600 permissions) to prevent instructions from being interpreted as agent commands.
[CREDENTIALS_UNSAFE]: The skill manages Google authentication state and session cookies locally. It follows security best practices by using os.umask(0o077) and explicitly setting restricted file permissions (0600 for files and 0700 for directories) to prevent unauthorized access to the browser profile and session data.

notebooklm