The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill requires the agent to access and process a local Google Workspace authentication file at /path/to/credentials.json using the gog auth credentials command.
[EXTERNAL_DOWNLOADS]: The setup instructions include downloading and installing the gogcli tool from a non-standard Homebrew tap (steipete/tap/gogcli) which is not on the trusted vendors list.
[COMMAND_EXECUTION]: The skill relies on executing system commands via the brew package manager and the gog CLI to perform operations across Google services.
[DATA_EXFILTRATION]: The skill facilitates the reading and transmission of sensitive data from Gmail, Google Drive, Sheets, and Docs. Specifically, the gog gmail send and gog drive download commands provide mechanisms to move data from the user's workspace to external recipients or local file paths.
[PERSISTENCE_MECHANISMS]: The skill documentation explicitly mentions using cron to poll emails every minute, which establishes a persistent execution path on the host system.
[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection as it ingests untrusted data from external sources (emails, files, and spreadsheets).
Ingestion points: Gmail message content, Drive file contents, and Sheets cell data are processed by the agent.
Boundary markers: No boundary markers or 'ignore' instructions are used when processing the retrieved content.
Capability inventory: The skill has broad capabilities including sending emails, creating calendar events, uploading files to Drive, and modifying spreadsheet data.
Sanitization: There is no evidence of sanitization or filtering applied to data retrieved from Google services before it is passed to the agent's context.

google