The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses string interpolation to build shell commands using variables like <PROMPT>, <title>, and <doc-path> (e.g., notebooklm generate slide-deck "<PROMPT>" ...). Since these variables can be sourced from user input or external documents without sanitization, a malicious string containing shell metacharacters such as backticks or semicolons could be used to execute arbitrary commands on the system.
[EXTERNAL_DOWNLOADS]: The skill requires installing notebooklm-py and playwright using uv tool install. These are third-party dependencies from public repositories that are not part of the standard agent environment.
[DATA_EXFILTRATION]: The core functionality involves uploading local document content to Google's NotebookLM service. While the skill includes a warning regarding sensitive documents, this process inherently transmits local data to an external cloud platform.
[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection because it processes content from external documents to populate command arguments.
Ingestion points: Local document content accessed in Workflow Step 1.
Boundary markers: Absent; content is extracted and used directly without delimiters or safety instructions.
Capability inventory: Execution of shell commands via the notebooklm CLI.
Sanitization: Absent; document-derived text is not escaped before being interpolated into shell commands.

wowerpoint