The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill establishes an attack surface for indirect prompt injection by combining tools for reading incoming email content with capabilities for sending or modifying emails.
Ingestion points: The skill utilizes gws gmail +watch and gws gmail users messages list to monitor and fetch incoming email data, including subjects and bodies from external senders.
Capability inventory: The agent is authorized to send emails (gws gmail +send), create persistent Gmail filters, and modify inbox labels.
Boundary markers: The instructions provide a security boundary by requiring the use of --dry-run and --draft flags for user review and explicitly directing the agent to treat all inbound content as data rather than instructions.
Sanitization: No programmatic sanitization or escaping of the email content is performed; safety depends on the model's adherence to the instructional constraints provided in the skill.
[EXTERNAL_DOWNLOADS]: The skill fetches and installs the @googleworkspace/cli package from the npm registry. This package is maintained by the googleworkspace organization, which is an established well-known service provider for Google Workspace tooling.
[COMMAND_EXECUTION]: The skill frequently executes shell commands via the gws CLI to interact with the Gmail API for authentication, email composition, and inbox management.

gws-hebrew-email-automation