The Agent Skills Directory

[COMMAND_EXECUTION]: The skill is designed around the execution of the gog command-line utility to interact with Google APIs. It provides exhaustive documentation for authentication (gog auth add), configuration, and service-specific commands. To ensure safe operation, the skill recommends global safety flags such as --dry-run for previewing changes, --no-input to prevent interactive hangs, and --gmail-no-send to block outgoing mail.
[INDIRECT_PROMPT_INJECTION]: The skill manages commands that ingest external data, creating a potential surface for indirect prompt injection.
Ingestion points: External content enters the agent's context through commands that read Gmail message bodies (gog gmail messages search --include-body), Google Docs text (gog docs cat), and file metadata from Google Drive.
Boundary markers: The documentation promotes the use of the --json flag to produce structured, machine-readable output, which facilitates the separation of data from instructions when processed by tools like jq.
Capability inventory: The gog utility possesses powerful capabilities, including sending emails, managing file sharing permissions (gog drive share), and executing remote logic via Apps Script (gog appscript run).
Sanitization: The skill mitigates risks by documenting environment-based restrictions such as GOG_ENABLE_COMMANDS, which allows users to explicitly whitelist only the necessary commands for a given task.

gog-cli