The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses subprocess.run to invoke the gws CLI when it is installed on the system. Analysis shows these calls involve static or well-defined arguments and are used as a legitimate execution backend for Google Workspace operations.
[DATA_EXPOSURE_AND_EXFILTRATION]: The skill explicitly instructs users to store their API credentials in .env or specific JSON files like google_token.json. These practices are consistent with standard secret management for local agent skills and do not represent a security risk.
[PROMPT_INJECTION]: The SKILL.md contains strict operational rules for the AI, such as 'Never send email... or modify Docs/Sheets without confirming with the user first'. These are safety guidelines designed to protect the user, not to override agent safety protocols.

google-workspace