postwall_skill

Overview

This skill provides a secure email gateway that places a human-in-the-loop between AI agents and email. It lets agents detect approved emails, read them only after human approval, and submit drafts for human review before sending. Use it to enforce approval workflows and prevent autonomous emailing.

How this skill works

The skill polls PostWall for unread emails that have been approved by a human reviewer and lists their metadata. Reading an email marks it as read so it won't be returned again. For outbound mail, the skill submits drafts to the PostWall dashboard and returns an approval URL; the message is only sent after a human approves it.

When to use it

• When agents must not read private email without explicit human approval
• When you need a human approval step before an AI sends email
• For polling inboxes for approved messages in automated workflows
• When you want to submit emails for review and share an approval link
• When you need structured JSON output for scripting and automation

Best practices

• Poll with postwall check every 5–10 minutes during active sessions to balance latency and rate
• Use postwall inbox --json to fetch metadata for filtering before reading full bodies
• Mark processed messages with postwall mark-read to avoid duplicate processing
• Submit drafts with postwall draft and share the approveUrl to track human approval
• Update drafts via postwall update only while status is pending; treat approved/rejected/sent as terminal

Example use cases

• Automated agent that only reads emails after a human flags them safe via the PostWall dashboard
• Batch metadata processing: filter inbox --json by subject or sender, then mark-read for processed IDs
• Workflow to draft customer responses: submit draft, send approval URL to manager, poll postwall status until approved
• Assistants that must never send emails autonomously can submit drafts and wait for human approval before delivery