​
Completed state

At the end of this guide, an active weekly operations digest has a saved Agent owner, project context, schedule, concurrency policy, and Chat output route. A manual test run has produced one linked conversation with inspectable Run evidence.

​
Prerequisites

Prepare a working Agent and a repeated request whose inputs and result shape are already stable. Decide who owns failures, what starts the work, and whether each result belongs in a new Chat or a tracked Issue. The exact output comparison is in Automation output routing.

​
Create the operations digest

  1. Open Automations and start a custom definition.
  2. Name it Weekly operations digest and select the reporting Agent and operations project.
  3. Write instructions that define the reporting window, source records, expected sections, and how to report missing data.
  4. Set the schedule to Monday at 09:00. The current UI has no timezone selector: it saves and shows the schedule in the browser’s local timezone.
  5. Keep Send to chat so every real execution gets a separate Messenger conversation. Choose tracked Issue output instead when the result needs explicit assignment, status, dependencies, or review.
  6. Select the concurrency behavior for an overlapping run. Choose skip or coalesce when duplicate reports would be harmful.
  7. Save the Automation, then dispatch it manually once.
  8. Open Run history and follow the linked Chat. Inspect the input message, Agent reply, source Agent Run, and terminal status.
Selecting a built-in template only fills an editable draft. It does not create an Automation until the definition is submitted.

​
Success signal

Automation detail shows the saved definition, next scheduled run, output mode, and test Run. The test produces one new Chat linked from that Run, and the Chat contains a useful result or visible failure evidence.

​
Recovery

If no work appears, inspect the Automation Run first. A skipped or coalesced record should name the concurrency reason and should not create an empty Chat. If dispatch failed after Chat output began, inspect the linked conversation and Agent Run for partial or fallback evidence. Correct the owner, context, schedule, runtime, or output route, then run one more manual test. Pause the Automation while repairing it so a schedule does not create additional attempts.