Catch a broken signup or checkout before your customers do

Every weekday at 6:45am and 1:45pm, walk my signup and checkout funnel in a real cloud browser and tell me it is broken before my customers find out. Use a single cron trigger that covers both run times (45 6,13 * * 1-5).

Start with Kernel and Create Browser Session. If my funnel begins behind a login, attach a saved profile (profile.name together with save_changes) so the session reuses stored login state instead of signing in from scratch on every run. Then call Start Browser Replay Recording and hold on to the returned replay ID for the rest of the run.

Drive the funnel with Kernel's Execute Batch Computer Actions, using Click Mouse in Browser, Type Text in Browser and Press Keys in Browser for the individual steps. Call Capture Browser Screenshot at each of these four checkpoints: (1) the pricing page loads, (2) the start trial button works and moves us on to signup, (3) the signup form submits, and (4) the post-signup dashboard or order confirmation actually appears. Keep each checkpoint's screenshot associated with its step so the evidence lines up with the verdict.

Always drive the flow with a dedicated test account and a test payment card so production data is never touched and no real charge is ever created. Never sign in as a live customer and never submit a real card number.

When the walk is finished, call Stop Browser Replay Recording and then Download Browser Replay, so there is a video of exactly what happened that I can link to.

Now judge each checkpoint on whether it genuinely succeeded, not on whether the page merely returned something that looks fine. A server can return a healthy page while the payment step silently fails. For each checkpoint, read the screenshot and confirm the specific expected element or state is really present: actual pricing tiers rendered rather than an empty grid, the trial button genuinely navigating onward, the signup form accepting the submission rather than reposting with a validation or server error, and the dashboard or order confirmation showing a signed-in state or a real order number. Treat spinners that never resolve, blank or skeleton states, error banners, and redirects back to the same page as failures even when the page itself loaded.

If a checkpoint failed, first use Linear Search Issues to look for an existing open issue about that same funnel step, so we never open a duplicate. If a matching open issue exists, use Add Comment to Issue to record this run against it. If there is no match, use Create Issue. Either way, include the failing step, what I expected to happen, what actually happened, the run timestamp, and the replay link.

Then use Slack Send a Message to post to my engineering channel with the broken step plus links to the Linear issue and the replay. Slack cannot take an uploaded file in this workflow, so carry the replay and any screenshot evidence as links or plain text rather than as attachments.

If every checkpoint passed, still send a short one line healthy confirmation to the same Slack channel, so silence is never mistaken for success and I can tell the difference between a passing funnel and a workflow that failed to run.