← Help

How do I read a flow run’s results?

Every time a Flow runs, its results appear on the flow’s page under Runs. Open a run to see what happened, step by step. A run that’s still going streams its steps in live as they complete.

Kinds of runs

• Dry run — a rehearsal you start from the editor. Approving it seeds the baseline and starts the schedule.
• Scheduled run — the automatic daily or weekly run.
• Manual run — started with Run now on the flow’s page. Limited to 10 per flow per day.
• Triggered run — started by a deploy trigger.

The verdict: new violations vs. the baseline

The headline of a completed run is the comparison against the flow’s baseline of known violations:

• New violations — issues that weren’t in the baseline. These are the regressions the flow exists to catch, listed with the rule, the element, and a screenshot of the page.
• No new violations — nothing to act on. The run may still note how many known violations remain in the baseline.
• Violations resolved — baseline issues that no longer appear. They’re marked resolved automatically; nothing to do.

You can dismiss a new violation right on the run report — it stays in the baseline and never re-alerts. See How do I manage known violations? for how dismissals work and how to reverse them. When a run has something worth knowing about, account members are alerted by email.

Reading the steps

Each step shows what it did, a screenshot of the page after it ran, its findings (split into new and known), and a status:

• OK — the step ran and found its target reliably.
• Fragile — the step ran, but only by falling back to a raw click on screen coordinates. The element isn’t in the page’s accessibility tree, so future runs can’t reliably lock onto it — and an element invisible to the accessibility tree is usually invisible to assistive technology too. The durable fix is in your markup: make the control a real button or link with an accessible name.
• Failed — the step couldn’t run. Later steps are skipped, since the journey can’t continue.
• Healed — the page changed since the last run (text, structure, or layout), the saved shortcut for the step no longer matched, and AccessLint re-found the target by its description and updated the shortcut. Healing is automatic; a healed step is working as designed.

A run also shows a durability rollup — how many of its steps matched, were fragile, or failed — so you can see a flow’s health at a glance.

Screenshots are kept for a week for routine runs, and indefinitely for notable ones (dry runs, failures, runs with new violations, healed or fragile steps).

Approving a dry run

A dry run ends with an approval prompt showing exactly what the baseline would be seeded with:

• If any step failed, the dry run can’t be approved. Fix the step and run another.
• If steps are fragile, you can still approve — but consider making them more robust first, or the flow may not hold up.
• Approving seeds the baseline from everything the run found and starts the schedule. From then on, only new violations alert.

When a run fails

• Couldn’t sign in — the run failed before any steps ran. The run page shows screenshots of each login phase so you can see how far it got, and the flow is paused so it doesn’t fail repeatedly; fix the domain’s authentication settings, then resume with a fresh run.
• Domain no longer verified — the run was stopped. Re-verify the domain.
• Something went wrong on our end — an engine error. The baseline is unchanged; the next run proceeds normally.