Audits everything in your account that starts or shows on its own—auto-starting flows and checklists, banners, and popup announcements—finds where users get too much at once or the wrong thing, and proposes one change for each item.
Download
What it does
The skill reads every start condition in full, however it's built, and works out which items can reach the same user at the same moment. It then checks that against what real users actually saw, by reading the event timelines of the people who got the most content, so you see real overload and not just what the rules predict. It also walks through a few typical users ("a signup from today", "a paying customer of a year") page by page.
Before suggesting anything, it works out what each item is for from its name, labels, conditions and what starts it, so follow-ups, onboarding sequences and recurring surveys aren't "fixed" by mistake. Every item then gets one action, with the evidence and a confidence level:
Keep—working as intended.
Fix—the smallest change that improves what users see, such as an interruption guard, a start limit or a corrected audience value.
Retire—unpublish it, with evidence that it's dead, stale or finished.
Review—a decision only your team can make, or a change the skill isn't sure enough about.
The skill checks every item before it reports back, however large your account is. On accounts with hundreds of items it works in batches and gives a short progress note as it goes ("checked 120 of 217 items") rather than stopping to ask whether to continue.
You get a short summary in chat first, then a decision tracker with Decision, Owner, Decision date and Applied columns for your team to fill in. Items that can start for the same user on the same page are put in overlap groups, and the tracker is sorted by them so related items sit together. At the start, the skill agrees with you where the tracker goes: a connected spreadsheet tool, which it checks works before the audit begins, or an .xlsx or CSV file.
What you'll need
The Userflow MCP connected
Editor access or above to apply fixes. With viewer access, the audit still runs as a read-only report, ending with a list of changes for an editor to apply.
How to use it
"Our users are seeing too many pop-ups. Audit our auto-start flows."
"Which flows show to trial users?"
"Clean up the old flows we don't need anymore."
By default it looks at all users over the last 30 days. Name an audience, such as trial users or a segment, and it marks which items target that audience and which reach it anyway.
What happens when you approve changes
Nothing changes until you pick which changes to apply and confirm the exact list. Then:
Fix changes are saved as drafts only. End users see nothing different until someone publishes the item, and whoever publishes it next ships these changes too.
Retire takes effect at once. Unpublishing stops the item for end users immediately. It isn't deletion: the item and its analytics stay in your account.
Some changes are handed to the Builder. Rules on pages, URLs, elements or segments can't be safely rewritten from the skill, so you get a Builder link and the exact change to make there.
The skill only publishes a fixed item if you ask it to, and shows you the impact before it does.
Things worth knowing
Settings are read from drafts. If an item has unpublished changes, what's live may differ from what the skill read. Those items are flagged, and fixes on them get a lower confidence.
Priority isn't the first thing Userflow checks. A flow the user has never seen starts before a higher-priority one they've already seen. The skill quotes the real start order rather than just suggesting "raise the priority".
The interruption guard only counts flows. It waits after the user's last flow, not after a checklist or banner. Banners and popup announcements don't wait in line at all; they can show at the same time as a flow.
A 0% completion rate isn't always bad news. Completion counts sessions that reached a goal step. A one-step flow with no goal step stays near 0% even when everyone sees it.
Recently published items aren't retired. Anything first published inside the window has partial data, so it gets Keep and a re-check instead.
Audience reach uses each user's attributes today, not what they were when the user saw the item.
If a tool keeps failing, you still get a summary. The skill tells you what failed and how many items it checked, and lists the unchecked items in the tracker as "not checked yet".
What to do with it
Share the tracker with whoever owns each item and let them fill in the Decision column. Publish the Fix drafts once they've been checked in the Builder. To keep things tidy as content grows, the userflow-optimize-labels skill gives every live item a consistent set of labels.