Deliverability Watchdog

Monitor email deliverability health and flag mailboxes or domains showing signs of spam or bounce issues before they damage sender reputation.

Instructions

When invoked (on-demand or recurring), query analytics for deliverability metrics across mailboxes and domains, identify anything trending in the wrong direction, and present a health report.

Steps

1. Confirm scope. Default to last 30 days, full account. Ask if the user wants a specific team, rep, or domain.

2. Gather the data by submitting these analytics questions via mcp__claude_ai_Amplemarket__ask_analytics:

   • Overall bounce rate and spam rate
   • Bounce rate and spam rate by mailbox
   • Bounce rate and spam rate by sending domain
   • Bounce rate and spam rate trend over the timeframe (weekly breakdown)
   • Open rate by mailbox (low open rate can signal spam folder placement)
   • Email volume by mailbox

   Submit all in parallel, poll with mcp__claude_ai_Amplemarket__get_analytics_result.

3. Analyze for problems. Look for:

   • Mailboxes or domains with elevated spam or bounce rates
   • Mailboxes with unusually low open rates compared to the account average (potential spam folder placement)
   • Trends -- is anything getting worse week over week?
   • Volume concentration -- any mailbox carrying a disproportionate share of sends?

4. Present a well-formatted health report. Lead with an overall health summary, then break down by mailbox and domain. Flag anything that needs attention and recommend actions.

   If running as a recurring Slack report, format using Slack mrkdwn and post to the configured channel. Only post if there's something worth flagging -- don't spam the channel when everything is healthy.

Examples

User prompt: "Check our email deliverability -- anything we should worry about?"

Simple example output: The agent finds overall bounce rate is 1.8% (healthy) but one mailbox has a 4.2% spam rate trending up over 3 weeks. Open rate on that mailbox dropped from 45% to 28%. Recommends pausing that mailbox, investigating content/authentication, and redistributing volume. Presented as a health summary table with per-mailbox breakdown and a trend section.

Troubleshooting