Failure triage and auto-heal
Investigate a failed run, group likely causes, and repair broken locators.
Start with the failed case's evidence. Decide whether the failure comes from the app, test expectations, access, data, or execution setup before changing the test.
Inspect the first failure
- Open Runs and select the execution.
- Filter to failed or needs-review cases.
- Open the first failing step and compare expected behavior with the screenshot and error.
- Check the account role, environment, and test data.
- Reproduce the affected workflow and record the cause.
Repeated failures can share an upstream cause, such as an expired login or an unavailable API.
Run AI triage
Use Triage where available to analyze failed cases and group likely causes. Wait for the analysis and review its conclusions against the evidence.
Triage uses generation capacity and the project's feature entitlements. Its explanation is a suggestion to investigate, not a confirmed application defect.
Repair a locator
After triage, use Auto-heal where available to inspect broken targets and update selectors. This changes the saved test steps.
Review the proposed or updated target and rerun the affected case. A locator that now resolves must still point to the intended control. Do not use auto-heal to remove a meaningful behavior assertion.
Verify the correction
Use Re-run failed only to recheck failed and needs-review cases, then run the related workflow if the fix affects more than one step.
Approval of a trusted test and the execution's needs-review state serve different purposes. See Create and edit tests.