The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses subprocess.run and subprocess.check_call in its scripts to manage the lifecycle of the Google Workspace integration.
scripts/setup.py uses subprocesses to check for and install required official Python packages (google-api-python-client, etc.) using pip or uv.
scripts/google_api.py and scripts/gws_bridge.py execute the gws CLI tool if it is available on the system path to perform Google Workspace operations.
These executions are restricted to the intended purpose of the skill and do not involve unsanitized user input flowing into shell commands.
[INDIRECT_PROMPT_INJECTION]: The skill processes external data from Gmail, Google Docs, and Google Drive which could contain malicious instructions.
Ingestion points: Reads email bodies, document content, and spreadsheet values (scripts/google_api.py).
Boundary markers: The instructions in SKILL.md advise the agent to confirm actions with the user, but do not specify explicit data delimiters.
Capability inventory: File system writes (drive download), network operations (Google API calls), and command execution (gws CLI).
Sanitization: Standard JSON parsing is used for API responses, but no specific NL sanitization is performed on ingested text before it enters the context.

google-workspace