Interrupt work

Replace the placeholders with values from your own authorized scope before copying this prompt into an MCP client.

Inputs

• {{workspace_id}} Stable workspace ID.
• {{application_id}} Stable application ID.
• {{task_id}} Exact task ID from the task record.

Prompt

Read task {{task_id}} in {{workspace_id}} / {{application_id}} with read_v2_task.
Check ownership or control authority, whether the task is interruptible, and which effects committed.
If authorized and interruptible, call interrupt_v2_task, then read the task again.
Report status, nativeStatus, committedEffectsRemain, and the final task event if present.
Stop on terminal state, denial, or unsupported interruption.

Example inputs

In the disposable Operating Data Cube, leave the export worker paused and request an export with facts.transfer.export.request. Use the member filter below, a fresh idempotencyKey, and an expiresAt Unix timestamp 15 minutes ahead. Read the application state first. Pass its commitRevision as expectedSnapshotRevision and its metadataRevision as expectedMetadataRevision.

{
  "Account": "InputA",
  "Entity": "PointA",
  "Year": "Y2027",
  "Period": "P01",
  "View": "Input",
  "Value": "Raw",
  "Custom01": "ItemA",
  "Scenario": "Actual",
  "DataType": "InputData"
}

Find the task whose sourceId equals the returned export requestId. An accepted export that has not started should report Interrupted, native Cancelled, and committedEffectsRemain: false after interruption. The task log retains the acceptance and interruption events. Use that same task for the rerun example.

Tools used include read_v2_program, execute_v2_source_program, list_v2_tasks, read_v2_task, interrupt_v2_task and read_v2_task_logs. See the module guidelines and tool index.