Black Friday falls on November 27 this year, and many teams are already asking their agents to prepare the holiday list. That part is quick. An agent can verify thousands of addresses and deliver a clean, deliverable-only file in a single afternoon.

The risk is that a faster route to a sending list can also be a faster route to a damaged domain. A deadline is a good reason to start early, but it’s never a reason to skip the checks that protect every campaign you send after the holidays.

That’s why your outbound agent needs one clear rule: a campaign moves into sending only when its verification, permission, suppression, sender setup, and volume checks all pass. If any evidence is missing, the campaign waits. Your agent becomes the checkpoint between an ambitious campaign and a sending decision your team can stand behind.

A Verified List Is the Starting Line, Not the Finish Line

Say your agent prepares a holiday offer for 10,000 contacts and builds a deliverable-only file. That’s useful work.

But did those contacts agree to receive this offer? Has anyone unsubscribed since the list was exported? Is your sending platform set up correctly? None of that shows up in a verification result.

A mailbox can accept your email while its owner wants nothing to do with your campaign. Google is explicit that subscription messages should only go to people who want them. Verification and permission answer different questions.

The Emailable CLI lets your agent verify addresses, but permission records, campaign controls, and sending reports have to come from the systems that own them.

So, the right handoff is “these addresses passed verification,” not “this campaign is safe to launch.”

Gmail, Yahoo, and Outlook Rules Still Apply When an Agent Does the Work

Automating campaign preparation doesn’t change what mailbox providers expect from senders, and the stakes keep rising. Since November 2025, Gmail has enforced its requirements more strictly, rejecting non-compliant mail outright instead of just delaying it.

Google defines a bulk sender as anyone sending 5,000 or more messages to personal Gmail accounts in 24 hours, counting everything from the same primary domain, including subdomains. Once a domain is classified as a bulk sender, that status is permanent.

Yahoo also has bulk-sender requirements but doesn’t publish a number, so don’t copy Gmail’s threshold into a Yahoo rule. Microsoft, meanwhile, has applied similar authentication requirements to senders of 5,000 or more messages a day to Outlook.com, Hotmail, and Live addresses since May 2025.

Staying under a threshold doesn’t get you off the hook, either. Gmail and Yahoo both require SPF or DKIM authentication from every sender, whatever the volume.

Before volume goes up, your agent’s review should cover three areas:

1. Authenticate the actual sending stream: Bulk senders need SPF and DKIM, a published DMARC policy of at least p=none, and domain alignment. In short, SPF lists the servers allowed to send for you, DKIM adds a verifiable signature, and DMARC ties both to the From address recipients see. Check real message results, not just whether the DNS records exist.

2. Make leaving easy: Marketing and subscription messages need a header-based unsubscribe and a visible link in the body. Gmail requires the RFC 8058 one-click method; Yahoo recommends it but also accepts mailto-based headers. Using RFC 8058 covers both, and unsubscribe requests should be honored within two days.

3. Check infrastructure, not just recipients: Valid forward and reverse DNS, standards-compliant formatting, and TLS encryption, which Gmail requires, all matter. A verification file can’t prove any of these sender-side conditions.

Your agent shouldn’t mark these complete just because someone configured the domain once.

Give Every Verification Result A Destination

For a conservative outbound workflow, start with this decision table. The actions are campaign-policy choices; they don’t redefine Emailable’s results.

Keep the state, reason, and flags like accept_all, role, and disposable as separate fields. An accept-all flag doesn’t override a firm deliverable or undeliverable result when Emailable returns one.

In our Claude Code test, a 77-address batch returned 36 deliverable, 38 risky, 1 undeliverable, and 2 unknown results. Under the policy above, 36 would move on to the remaining checks, and the other 41 would stay out of sending until someone reviewed them.

That’s an illustration of routing, not proof that 41 bounces were prevented. The test covered verification and file handling, not campaign performance.

Check Before Every Send Without Repeating Work

Make verification a requirement before every campaign release, follow-ups included. Then decide what counts as current enough for that send.

Your written policy should spell out when fresh verification is required and when an existing result can be reused. New or changed addresses always need their own check, and any record missing a verification timestamp goes to review.

Say your agent already ran an approved verification for Tuesday’s campaign. Creating three exports from that result shouldn’t trigger three new batches. But if the campaign slips to next month, it needs a fresh look.

Use the actual verification time or batch record, not the CSV’s download date. Save the batch ID so an interrupted job can resume without resubmitting the list, and if verification becomes unavailable or the credit budget runs out, hold the unchecked contacts rather than guessing.

Apply current unsubscribes and do-not-contact records right before release, too. If a contact unsubscribes after Tuesday’s file is built, their address may still be deliverable, but Wednesday’s follow-up should leave them out. A deliverable result never overrides a suppression.

Let Sending Capacity Set the Pace

A bigger verified file isn’t a reason to send it all at once.

Google recommends sending consistently rather than in bursts, increasing volume gradually, and slowing down when bounces or temporary deferrals rise. There’s no universal daily number that keeps every domain safe; your sending history and recipient feedback decide that.

Take a team that normally sends 800 emails a day. Its agent prepares 8,000 eligible contacts for a holiday promotion. The right next step is a reviewed sending plan, not a tenfold jump just because the list is ready.

Set the pace in your sending platform, where you can actually enforce it, and coordinate every agent and campaign sharing the same infrastructure. Five agents each respecting their own limits can still blow past the team’s overall allowance.

Keep promotional mail separate from password resets and receipts by using different sending addresses, as Google recommends, though that won’t change your primary domain’s bulk-sender classification.

Your agent prepares the queue and explains the constraints. Your sending system enforces them.

Treat Complaints As Feedback, Not A Rounding Error

Google recommends keeping your spam rate below 0.1% and never reaching 0.3% or higher. Yahoo requires it to stay below 0.3%. Those numbers aren’t a complaint budget to spend.

Read each provider’s own metric, too. Gmail’s Postmaster Tools and Yahoo both base their rates on inbox-delivered mail, so dividing your complaints by every address in your campaign won’t reproduce either number.

A low reported rate can also sit alongside poor delivery when lots of messages already land in spam, and missing dashboard data isn’t proof of zero complaints.

Give your agent clear pause conditions: rising complaints, unusual deferrals, authentication failures, or a volume increase nobody approved. If monitoring is delayed or unavailable, the agent shouldn’t approve more volume assuming everything’s fine.

If your DKIM domain is enrolled in Yahoo’s complaint feedback loop, you’ll get reports that help you suppress complaining recipients from future campaigns. Feed those back into the next pre-send review.

A verified address never outranks a recipient’s response.

Build A Pre-Send Gate Your Team Will Actually Use

A pre-send gate is a repeatable decision before release, not another spreadsheet someone forgets to read. Here’s how to turn the checks above into an agent skill:

1. Trigger it at release: Run the gate before any campaign or follow-up enters sending, and run it again whenever the audience, send time, message, or sending setup changes significantly.

2. Require the right inputs: The audience, verification results and timestamps, permission and suppression records, sender evidence, delivery reports, and sending limits. If one is missing, the campaign holds.

3. Apply explicit decisions: Use the result table, remove contacts who can’t be emailed, and keep unresolved records separate. Check campaign-wide requirements on their own, too, because a clean audience can’t make up for broken authentication or unsubscribe handling.

4. Return an output anyone can follow: Produce a candidate file, an exclusion-and-review file, and a release report with counts, reasons, evidence dates, the rule version, and the responsible owner. Reconcile the totals so no contact goes missing.

5. Enforce the boundary: Have the sending system reject any release without a valid gate decision and human approval for that exact campaign version. Limit verification spending and cap retries.

6. Test the hold as well as the pass: Run an authorized real list as a dry run with sending turned off. Then, in labeled test copies, remove required evidence and add a suppression match. The gate should hold the affected record or campaign.

Once the rules and supporting records are in place, the request is straightforward:

“Review this campaign against our pre-send rules and explain what still blocks it. Don’t send anything.”

If unsubscribe evidence is missing, the right output is a held campaign that names that gap, not a vague “looks good.”

One honest limit: until the gate is wired into your release process, it’s a review report. A prompt can’t block a separate sending system on its own.

Scale The Decision, Not Just The Volume

The teams that come out of the holidays with their domain reputation intact won’t be the ones that sent the most. They’ll be the ones whose agents knew when to move forward and when to stop.

Start with one campaign. Let your agent verify, separate, and explain, make release controls enforceable, and let real sending feedback guide the next increase.

That’s useful autonomy: more preparation handled consistently, without turning a bigger list into an automatic instruction to send.