What problem does it solve?
It prevents unauthorized or accidental Gmail operations by requiring explicit user requests or database-owned connector action policies before sending, replying, labeling, triaging spam, or deleting messages.
Core Features & Use Cases
- Outbound Messaging: Send new emails and reply to existing threads using gmail_send and gmail_reply with explicit recipients, subjects, and bodies.
- Mailbox Organization: Apply or remove labels and triage spam by marking messages as spam, not-spam, or archiving them.
- Policy-Gated Autonomy: Heartbeat-triggered actions only run when a matching connector action policy exists in the database; deletions are Trash-only by default and never autonomous.
- Use Case: A user asks the agent to reply to a specific thread and archive a spam message; the skill verifies the request is explicit in the current conversation, then calls gmail_reply and gmail_spam_triage without escalating beyond the stated intent.
Quick Start
Ask the agent to send an email to a specific recipient with a subject and body, or to reply to a named thread, after connecting your Gmail account.