Investigate a failure

1. Start at Home and open an attention item, or select a project and open Issues. A failed query is shown as unavailable; retry it before concluding that a project has no failures.
2. Select Ongoing to focus on fingerprints that still have open failures. Expand a row to see when the issue first and last appeared and its complete fingerprint. The occurrence count and recovered count describe the observed history for that fingerprint.
3. Use Search history for this task to open the task list. This searches the sample task name and can include other failures and successful outcomes; it is not an exact fingerprint filter. Refine state and priority as needed.
4. Open a task. Read its recorded exception and traceback, execution timing and timeline. Expand arguments or event payloads only when they help the investigation. Redaction may remove sensitive values from captured data.
5. If a corrective action is appropriate, select the agent that owns the task. The available controls depend on your role and the agent's advertised capabilities. Confirm the result in Commands and inspect the subsequent task activity. A submitted command is not a completed action.
6. Return with Back to tasks to keep the list filters. For recurring problems, consider a dry-run automation rule and review its firings before arming it.

Try this flow with the sample incident. The demo blocks mutations; it does not run a real retry.