The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs the user to install notebooklm-py using uv tool install. This is an unofficial, third-party package not hosted by a well-known or trusted organization.
[COMMAND_EXECUTION]: The skill executes multiple shell commands via the notebooklm CLI to manage notebooks, upload files, and download artifacts. These commands interact directly with the host filesystem and external network services.
[INDIRECT_PROMPT_INJECTION]:
Ingestion points: The skill ingests untrusted external data through notebooklm source add (URLs, YouTube videos, PDFs, etc.) and notebooklm source add-research (web/drive research).
Boundary markers: There are no explicit instructions or delimiters provided to prevent the agent from obeying malicious instructions embedded within the imported sources.
Capability inventory: The skill possesses capabilities to write files to the local system (notebooklm download), perform network operations, and generate content based on source data.
Sanitization: No sanitization or validation logic is evident for content retrieved from external research or uploaded files before it is processed by the agent.
[CREDENTIALS_UNSAFE]: The skill manages sensitive Google OAuth credentials. It prompts for notebooklm login and stores authentication state in ~/.notebooklm/context.json or accepts sensitive NOTEBOOKLM_AUTH_JSON environment variables, which could be exposed if the environment is compromised.

notebooklm