Changelog
The agents diagnose like a consultant, from a published playbook
The agents now work a problem the way our own team does, from a published playbook: what to ask, the checks cheapest first, what our studies measured each finding to cost, and when new inboxes are and are not the answer. The most-used checks come back with that reading attached, and a SURBL listing is reported for what our data says it costs rather than as a broken domain.
For AI agents
- A new tool, get_diagnostic_playbook, turns a symptom in the person’s words ("reply rate dropped", "listed on SURBL", "SCL 5") into the questions to ask, the checks to run cheapest first, the studies and guides to read with full URLs, and the evidence that would make new domains or inboxes the right fix. Eight scenarios, on both agents. the playbook
- A reply-rate drop is worked in order: setup, then lower sending volume and more warmup, then the bounces, then the list and the recipients behind security gateways, then sender-recipient routing, then the copy, and only then replacing the domains the evidence condemns. the reply-rate scenario
- The most-used checks (SURBL, blacklists, reputation, the deliverability grade, bounces, Outlook headers, copy and ESP segmentation) now return a guidance block: what the result means in proportion, the next steps, and the studies behind it, quoted from the studies’ own data. the SURBL study
- The hand-off that called infrastructure degraded no longer fires on a blacklist listing of a domain’s website host, a minor list, or a failing grade made of missing DNS records; only on a major list naming the sending address, a failing grade whose failure is the blocklists, a domain-level listing or a burned inbox. the Deliverability Agent
Pages
- The playbook is published in full at /plugins/playbook, word for word what the agents follow, and linked from both agent pages. /plugins/playbook