Coordinate and recover work
Use this guide when you need to coordinate or recover work in Work & Settings. End with a checked Task state and a result its owner can verify.
Before you start
A work operator needs the Task, its intended result, and access to the object it concerns. Start from the saved Task if one exists. Use Chat for discussion. Create a Task only when a durable assignment is needed. Confirm who can own the result and which actions its current lifecycle allows. A notification is a pointer to that work, not evidence that the work finished.
1. Assign the work
Coordinate work from the Console Task collection at /tasks. Create or open the Task and assign an eligible principal. Set the due time when a deadline matters. Read the saved owner, state, and deadline back before relying on them. The scheduling contract owns the exact call. Omitting the due time clears the deadline rather than preserving it.
2. Record the result
Record work context as a Task comment, progress entry, or attached evidence. The Task's local history records that change with the assignment. Read the evidence before selecting an allowed state transition. A comment does not change the state. If the Task concerns a Run, open that Run for its result. Use Activity when the evidence crosses execution kinds.
3. Recover missed work
Recover work by opening the Task named by a late or missing notice. Read its state, owner, deadline, and history. If work remains, agree on the next action and revise the assignment or deadline as needed. For failed execution, follow the linked service's recovery path. A review Task needs its owning approval decision. Marking a notice read is not a Task repair.
Checkpoints
Work has a checkpoint after each change: the saved Task names its owner, the deadline matches the plan, and history contains the reported evidence. After a transition, read the current state again. If Activity is partial or unavailable, inspect the owning execution before claiming its result. A notification's read state cannot fill an evidence gap in either the Task or Run.
Success
Work coordination succeeds when the Task has an eligible owner, a checked state, and evidence for the next action or finished result. Recovery succeeds when the underlying assignment or execution shows the intended change. A later deadline alone does not prove that missed work was done. Keep the Task's history available so another operator can see what changed and what still remains.
When it fails
Work changes can be refused even when the Task is visible. Read the exact cause before submitting another change.
TASK_STATE_STALE
The Task changed after the read used for your transition. Reload the Task, inspect the latest history, and choose from the allowed actions again. Do not overwrite another worker's progress with an old state.
TASK_TRANSITION_NOT_SUPPORTED
The requested action is not valid for the current lifecycle state. Choose an allowed move. A terminal Task requires the dedicated reopen action, which still checks its pinned lifecycle revision.
TASK_PRINCIPAL_NOT_ASSIGNABLE
The chosen principal cannot take this Task. Select an eligible owner and read the saved assignment back. Before an Agent executes the Task, it must meet its Taskable rules.
Recovery
Work recovery and notice recovery need separate checks. Open Notifications and settings to inspect disabled channels or digest timing when delivery is missing. Save the intended preferences and read them back. For a degraded review notice, use Reviews and repair to restore delivery from the waiting gate. The review Task remains usable while notification delivery needs repair.