How to use this
This is a working library of Claude prompts for optimizing your store's Klaviyo deliverability.
They're organized the way deliverability work actually goes: diagnose what's happening on your list, recover the placement you've lost, and maintain it so it doesn't slip again. No tricks, just the systematic moves that keep your email landing in the inbox instead of the Promotions tab or spam.
Every prompt follows the same shape:
- a one-line use case so you know when to reach for it
- the prompt itself ready to paste (with the output format built in)
- an operator note flagging what to watch for
Diagnose
Find what's hurting your inbox placement
Recover
Fix lost placement and clean dead weight
Maintain
Keep your list and sender reputation healthy
Get the full playbook.
18 copy-paste prompts, the Cowork + Tasklet scheduling setup, and the bonus stress-test for paid data providers. Built for brands running Klaviyo on their own store.
- 18 production-grade prompts with built-in output formats
- Scheduling setup, Cowork for hands-off monitoring, Tasklet when you want more
- Bonus prompt: the test a data provider should be able to pass
Before you start:
- Connect your Klaviyo account. These prompts read whatever account the Klaviyo connector is pointed at. Connect your store, and Claude works against your real data, every number it returns is yours, not a generic benchmark.
- Adjust everything bracketed, and keep each prompt's "Format the result like this" block, it's what gives you clean, copy-pasteable output instead of conversational prose.
- Read before approving any write. The diagnostic prompts are read-only. The few that suppress or subscribe (clearly marked) show the plan first, read it before you confirm.
- Ask for a chart. Several prompts end with a chart request, Claude renders the data as an HTML chart you can screenshot. Any data prompt works this way: after the result, add "now build that as an HTML chart."
- Your own numbers are the benchmark. A 28% open rate is healthy for one store and an alarm for another. Several prompts here establish your own normal first, then measure against it, which is the only comparison that actually means anything.
- Put the recurring ones on a schedule. The monitoring prompts are built to run on a cadence, and Claude Cowork (desktop) can run them automatically. Where a prompt is worth scheduling, a suggested cadence appears right after it. Full setup is in "Scheduling in Cowork" just below.
Scheduling these in Claude Cowork
The diagnostics and monitors earn their keep when they run on a cadence, your weekly pulse ready when you sit down Monday, a monthly scorecard prepped on the 1st. Claude Cowork (the desktop app) does this. You set a prompt up once, pick how often it runs, and Cowork runs it on your connected Klaviyo account, saving the output each time. Setup is about two minutes per task.
Two ways to set one up.
The quick way, from a chat: open Cowork, start a new task, paste the prompt you want to recur, then type /schedule and send. Claude asks a few multiple-choice questions, how often, what time, where to save, then confirms the task, its cadence, and what it does.
The precise way, from the sidebar: click Scheduled, then New task, and enter the prompt, frequency, and working folder directly in the form. Since the prompts here are already well-defined, you're just pasting a finished one and setting a cadence.
What to add when you schedule. A good recurring prompt also says where to put the result, e.g. "save the result as a dated file, deliverability-pulse-[date].md." That dated trail is half the value when you're comparing month over month.
A few things worth knowing.
- After the first run, Claude tidies your prompt into standing instructions based on what it learned, so later runs are sharper.
- Every scheduled task lives in the Scheduled tab, view upcoming and past runs, open any result, edit the prompt or cadence, pause, or delete.
- Claude Desktop notifies you when a run finishes.
- Scheduled tasks only run while your computer is awake and Claude Desktop is open. If the machine's asleep at the scheduled time, Cowork runs the task next time you open the app.
Where a prompt below is worth automating, you'll see a Schedule it line with a suggested cadence. The rest are run-when-you-need-them.
Scheduling that runs itself: a dedicated cloud agent
Cowork scheduling works well, with one catch: it only runs while your computer is awake and the desktop app is open. For a weekly pulse that's usually fine. But if you want your monitoring to run truly hands-off, firing on time whether or not your laptop is on, or if you run more than one store and don't want to reconnect Klaviyo every time you switch between them, there's a cleaner setup.
Give your monitoring its own dedicated agent that stays permanently connected to your Klaviyo account and runs in the cloud. We do this with Tasklet.ai, a cloud agent platform (from the team behind Shortwave) that connects to MCP servers, including Klaviyo's, and runs tasks on a schedule 24/7, without your machine being on. Because the agent holds its own connection, your scheduled reports just run, and if you own several stores, you give each one its own agent so there's never any connection-switching.
Setting it up:
- Create a Tasklet account and, in the dashboard, add the Klaviyo MCP server as a connection (Tasklet supports custom MCP servers alongside its prebuilt integrations). You authorize Klaviyo the same way you do in Claude, the difference is where the connection lives.
- One agent per store. If you run a single store, that's one agent permanently connected to your account. If you run several, set up a separate agent per store, each connected to its own Klaviyo, name them clearly so they don't get crossed.
- Give it the prompts in plain English. Paste the monitoring prompts from this guide (the pulse, the anomaly scan, the monthly scorecard) and tell the agent when to run them.
- Set the schedule and let it run in the cloud. Results land on the cadence you set, with no machine to keep awake.
A couple of honest caveats: Tasklet is a paid service, so it's worth it once truly hands-off scheduling (or running multiple stores) saves you more than it costs, for a single store checked weekly, Cowork is free and fine. It's also a newer platform that's still maturing, so connect it with least-privilege permissions where Klaviyo allows, and review the first few runs before you rely on them.
00SEGMENT SETUP, DO THIS FIRST
Several prompts below depend on segments existing in the Klaviyo account. Build these once per account before running any prompt that requires them. Name them exactly as shown so the prompts can find them.
- From the left nav, go to Audience → Lists & Segments, then click Create List / Segment at the top right and choose Segment.
- Name the segment exactly as shown below (e.g. Active) so the prompts can find it by name.
- Build the definition using the "What someone has done (or not done)" condition. Pick the metric (Opened Email, Clicked Email, Placed Order), the timeframe ("in the last N days"), and the count ("at least once" or "zero times").
- Apple Privacy Open = false must be added to every Opened Email condition individually. Under each Opened Email line, click the small + Add filter link and add Apple Privacy Open = false. If you have two Opened Email conditions in the same segment (e.g. Cooling, which has both "Opened = 0 in 30d" and "Opened ≥ 1 in 90d"), you need to add this filter to both, not just the first one. This excludes Apple Mail's auto-opens, which otherwise inflate engaged cohorts and shrink Dead.
- Chain conditions with AND. Click Create Segment when done. Klaviyo takes 10-30 minutes to populate the profile count on a list of any size, the segment will show "Processing" until it's ready.
The segment names above are what the prompts look for. If existing segments match these definitions but use different names, either rename them or update the prompts when you run them.
01Diagnose
01The 30-second health read
A quick, regular read of your core deliverability signals over time.
Schedule it: in Cowork, paste this prompt and `/schedule` it to run weekly. A standing Monday read keeps the baseline picture current.
Operator note
This is the read to run on a regular cadence, not just when something feels off, its value is the baseline picture it builds over time. Weekly buckets matter because a monthly average smooths over the week-to-week movement where problems first show up. When a week does stand out, the next question is whether it's a one-off (a heavy send, a single off campaign) or the start of a trend; one soft week is noise, three in the same direction is a signal. The rest of this section is how you dig into anything this read surfaces.
02Establish your baseline
Before you can spot trouble, you need to know your own normal. This is the reference every other prompt measures against.
Schedule it: in Cowork, paste this prompt and `/schedule` it to run quarterly. Re-establish your baseline as the list and program mature, then feed it to the monitoring prompts.
Operator note
This is the most important prompt in the book and the one people skip. "Is a 28% open rate bad?" has no universal answer, it depends entirely on whether your list normally runs 32% or 22%. A generic benchmark will either scare you about a healthy list or reassure you about a sick one. Save this output and re-run it quarterly; your baseline shifts as your list and program mature, and an alert built on last year's normal is worse than no alert. Every "is this bad?" question downstream gets answered by comparing to this, not to an industry chart.
03Deliverability by mailbox provider
"My open rate dropped" means nothing. "Gmail dropped and Yahoo didn't" means everything.
Operator note
Gmail and Yahoo run engagement-based placement: they watch whether your recipients open, and demote you when they don't. Because Gmail is usually the majority of a list, a Gmail-specific dip drags your blended number down and hides what's actually fine elsewhere. If one provider is bleeding and the others are steady, you have a placement problem at that provider, not a list-wide content problem.
04Inbox placement read
Engagement tells you if they opened. Placement tells you if they ever saw it.
Operator note
The connector reads engagement, not folder placement, so this prompt infers placement from the gap between a provider's open rate and your baseline. A big Gmail-specific gap almost always means Promotions-tab or spam-folder placement, because the email arrived but fewer people saw it. It's a strong signal but not proof; the only way to confirm folder placement is a seed-list test. Use this prompt to decide whether a seed test is worth setting up, not as a replacement for one.
05Is the dead weight measurable?
Quantify how much your disengaged subscribers are costing your placement.
Operator note
This is the single most common deliverability problem in ecommerce Klaviyo accounts and the easiest to fix. Mailbox providers judge you on engagement density, the share of recipients who actually want your mail. A third of the list opening at 3% isn't neutral; it's actively training Gmail to file you under Promotions or worse. The fix isn't "delete them today", it's a sunset flow that gives them one last chance and then stops mailing them.
06Authentication and infrastructure check
The boring stuff that silently caps everything else.
Operator note
DMARC at p=none is the most common gap. It passes the technical check, the record exists, so people assume they're done, but none means "monitor and do nothing," which doesn't satisfy Gmail/Yahoo bulk sender enforcement. Move to quarantine, watch the aggregate reports for a few weeks to confirm nothing legitimate is failing, then go to reject. Never jump straight to reject on a domain you haven't been monitoring.
07Is over-mailing the cause?
The most common self-inflicted deliverability wound: sending more than your list wants.
Operator note
Over-mailing is the deliverability problem people cause while trying to hit a revenue number, and it's invisible if you only look at monthly averages, you have to line volume up against engagement week by week. When opens fall and complaints rise as you send more, your list is telling you its ceiling. The fix isn't always "send less", it's often "send the same volume to a tighter, more engaged audience."
02Recover
08Build the suppression candidate list
Decide who to stop mailing, carefully, with the plan shown first.
Operator note
The protections are the whole point of doing this through a checklist instead of a raw filter. Someone who bought last month but hasn't opened an email is not a suppression candidate, they engage through your site, not your inbox. Pull those out before you do anything. And split the candidates: a 200-day-quiet subscriber deserves a sunset attempt; a three-years-quiet one should just go. Suppressing in Klaviyo doesn't delete the profile, it stops you mailing them.
09Find the complaint sources
Spam complaints are the fastest way to get throttled. Trace where they come from.
Operator note
Complaints cluster around two things: aggressive promotional sends to broad audiences, and badly-acquired subscribers. A giveaway list complaining at 0.9% is a textbook example, those people wanted the prize, never wanted you, and hitting "spam" is easier than unsubscribing. The fix is upstream: either stop importing giveaway lists into your main audience, or route them through a confirmation step first.
10Find your repeat complainers
One person complaining twice is a louder deliverability signal than the rate alone. Find them and cut them.
Spam-complaint rate tells you how big the problem is; this tells you who. A single complaint can be a misclick or a forgotten signup, but the same address complaining twice is unambiguous, they do not want your mail and every future send to them is pure downside. Suppress repeat complainers on sight.
11Audit the welcome flow for deliverability
Your first impression sets your inbox placement with every new subscriber.
Operator note
The welcome flow is where you teach mailbox providers whether new subscribers want you. A high first-email open rate is the best deliverability signal you have, it tells Gmail "this sender's new subscribers engage immediately." The trap is the second or third email that nobody asked for; a brand-story email complaining at triple the rate of the others is quietly poisoning the impression. Cut or replace the weak link before it costs you placement on everything else.
12Rebuild reputation after a cleanup
Once the list is clean, ramp back carefully. Don't undo the recovery with one big blast.
Operator note
After a cleanup or any sending break, mailbox providers re-evaluate you, and the fastest way to waste a good cleanup is to immediately blast your widest audience. Start narrow with your most engaged subscribers (the people guaranteed to open), let those strong signals rebuild your reputation, then widen. The go/no-go gates are tied to your own baseline, not a generic target, because "safe to widen" means "still performing like my healthy normal."
03Maintain
13Pre-send deliverability gate
Run before every broadcast. Catches the things that quietly tank a send.
Operator note
The line that matters here is "engaged share of audience." Mailing "All Subscribers" feels like reach but it's the fastest way to erode placement, because every cold subscriber who ignores the send tells Gmail you're sending unwanted mail. As a rule, broad promotional blasts should go to engaged-90-day; only your best content to the wider list, and only occasionally.
14Weekly deliverability pulse
A 60-second Monday read so drift never surprises you.
Schedule it: in Cowork, paste this prompt and `/schedule` it to run weekly. This is the canonical one to automate, a Monday pulse you read over coffee.
Operator note
Deliverability problems are almost never sudden, they announce themselves for weeks before placement actually drops, in exactly these five numbers. A single week of movement isn't a verdict, but three consecutive weeks of complaints and unsubs creeping up while opens slide is a list-quality problem forming in real time.
15What's abnormal for me?
Stop comparing to industry averages. Compare to your own normal and surface only the real deviations.
Schedule it: in Cowork, paste this prompt and `/schedule` it to run weekly. It surfaces only what fell outside your normal range, so a clean week stays quiet.
Operator note
This is the baseline idea turned into a working alarm. The point of establishing your normal range isn't to admire it, it's so a prompt like this can ignore everything sitting in-band and surface only what genuinely changed for you.
16Monthly list-quality scorecard
One number you can track and a founder can understand.
Schedule it: in Cowork, paste this prompt and `/schedule` it to run monthly. Run it on the 1st so each month opens with a fresh scorecard.
Operator note
"Engaged share" is the one number to put on a wall. It's the cleanest single proxy for whether mailbox providers consider you a wanted sender, and it moves slowly enough that a real change means something.
17The quarterly deep audit
The full picture, once a quarter, so nothing compounds in the dark.
Schedule it: in Cowork, paste this prompt and `/schedule` it to run quarterly. Put it on the calendar, this is the audit that compounds when skipped.
Operator note
Deliverability is the one area where doing nothing actively makes things worse, list quality only decays, never improves on its own. The version of your list you have today is the healthiest it will ever be without maintenance.
18Is your list growth actually healthy?
Adding subscribers isn't growth if the new ones never open. Check the quality of what's coming in.
Operator note
Most brands measure list growth as a single number going up, which hides the most important fact: where the growth comes from determines whether it helps or hurts. Fix acquisition upstream and you stop manufacturing the problem the rest of this book exists to clean up.
19Build a deliverability dashboard
One visual that puts the whole picture on a screen, for you or to hand to a client.
Schedule it: in Cowork, paste this prompt and `/schedule` it to run monthly. A standing dashboard run gives you a visual history without a monitoring tool.
Operator note
This is the prompt that turns everything else in the book into something you can look at in five seconds. The connector gives Claude the data; the dashboard makes it legible. The detail that makes it useful is "against my baseline": a generic dashboard shows numbers, but one that shades your normal band and marks your danger line tells you whether to act.
★Put your data provider to the test
If you buy or enrich contacts from a third-party data or identity provider, here's the test that source should be able to pass.
Operator note
Read the gap, not the absolute. A provider's contacts opening a few points under your list average is normal, new contacts haven't warmed up yet. A cohort running multiples of your normal bounce or spam rate is the red flag: those contacts are actively eroding your sender reputation, and you're paying for the privilege. Bounce and complaint rates are the truest tells, because they reflect address validity and genuine consent.
Next steps
Two ways to take this further.
A blood test for your Klaviyo account. It goes beyond what these prompts surface, benchmarking your account against other merchants of the same size and industry, so you see exactly where you're outperforming and where you need work.
Grade my account →Moves your email out of the Gmail Promotions tab and into the primary inbox, a direct lift in placement, and the engagement that follows, for very little work. A one-time $99 tool with an immediate impact on how your account performs.
Improve placement →By CustomersAI · customers.ai · First in a series for brands running Klaviyo inside Claude.