The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes subprocess.run to manage its internal environment and execute its component scripts. This is primarily seen in scripts/run.py, which acts as a wrapper for executing other scripts in the virtual environment, and scripts/setup_environment.py, which handles dependency installation. The implementation includes checks to ensure that the target scripts exist within the skill's directory.
[EXTERNAL_DOWNLOADS]: The skill downloads browser binaries required for its automation tasks. Specifically, scripts/setup_environment.py executes patchright install chrome to set up the Google Chrome browser. These downloads originate from well-known sources associated with browser distribution and automation tools.
[INDIRECT_PROMPT_INJECTION]: The skill's primary function involves retrieving content from Google NotebookLM and providing it to the agent. This represents an indirect prompt injection attack surface where untrusted data from documents or web pages could contain malicious instructions.
Ingestion points: Content is extracted from the NotebookLM web interface in scripts/ask_question.py using CSS selectors such as .to-user-container .message-text-content.
Boundary markers: The skill does not employ explicit boundary markers or delimiters to separate retrieved content from system instructions when presenting it to the agent.
Capability inventory: The skill has the ability to read and write local files (e.g., library.json, auth state) and execute shell commands through its orchestration scripts.
Sanitization: There is no evidence of sanitization or filtering applied to the text retrieved from the browser before it is passed to the AI agent.

notebooklm