The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill requires the installation of notebooklm-py and playwright. While playwright is a well-known service from a trusted organization, notebooklm-py is an external dependency from a source not included in the trusted vendors list.\n- [CREDENTIALS_UNSAFE]: The skill manages authentication by storing Google/NotebookLM session cookies in ~/.notebooklm/storage_state.json. These are sensitive credentials that allow access to the user's NotebookLM account.\n- [COMMAND_EXECUTION]: The script scripts/import_sources.py uses subprocess.run to execute the notebooklm CLI tool, utilizing source_id values parsed from external JSON files as command arguments.\n- [PROMPT_INJECTION]: The skill has a surface for indirect prompt injection as it imports external summaries and Q&A answers from NotebookLM into the vault.\n
Ingestion points: Data is ingested through sources.json and qa-output.json in the scripts import_sources.py and resolve_citations.py.\n
Boundary markers: No explicit delimiters or instructions are used to prevent the agent from obeying instructions embedded in the imported text.\n
Capability inventory: The skill can execute subprocesses via the notebooklm CLI and perform file write operations across the vault.\n
Sanitization: Filenames are sanitized using regex, but the imported text content is written to Markdown files without filtering for potential malicious instructions.

notebooklm