deepline-agentmail-inbox

name	deepline-agentmail-inbox
description	Use for Deepline GTM AgentMail workflows: inbox triage, inbound context capture, reply drafting, thread summaries, and safe email identity management. Draft-only by default.

Deepline AgentMail Inbox Skill

Trigger

Use this when the user asks for:

• checking the agent inbox
• summarizing inbound email threads
• drafting a reply from the Deepline GTM agent
• routing inbound requests into Deepline workflows
• creating or verifying an agent-owned inbox

Required Context

Read first:

• context/deepline_gtm_context.md
• context/claims_and_exclusions.md
• context/jai_voice_and_copy_rules.md
• rules/agent_operating_rules.md
• hermes/agentmail_primary_inbox.md

Rules

• AgentMail is an identity and inbound routing channel, not an unsupervised outbound system.
• Never send or schedule email without explicit approval.
• Prefer draft creation over send.
• Use Deepline for GTM data execution, enrichment, logs, usage, and output lineage.
• Use AgentMail only for inbox/thread/message/draft operations.
• Never expose AGENTMAIL_API_KEY in chat, Markdown, logs, or committed files.

Workflow

1. Identify the inbox job: list inboxes, read thread, summarize, draft reply, or route work.
2. Confirm AGENTMAIL_INBOX_ID and AGENTMAIL_EMAIL are configured.
3. Read only the relevant thread/message.
4. If GTM data is needed, start a Deepline Session UI plan and use Deepline-native tools.
5. Draft the response.
6. Add claims that need review.
7. Ask for explicit approval before send or scheduling.

Output Pattern

Thread summary:

• sender
• ask
• context
• suggested owner
• recommended next action
• risk / approval note

Reply draft:

• subject
• body
• source context
• claims needing review
• send approval required