Ablaufausführung verfolgen

Execution Tracking & Verification

Use this skill when building workflow run inspection UIs, step execution timelines, cancellation flows, and end-to-end verification tests for workflow systems.

Traceable Run, Step & Attempt Hierarchy

1. Three-Level Execution Identity (runId -> stepExecutionId -> attemptId):
   • Record every workflow run (runId, workflowVersionId, tenantId, triggerEventId, status, timestamps), every node activation (stepExecutionId, nodeId, status, resolved inputs/outputs), and every individual retry attempt (attemptId, attemptNumber, error details, duration, idempotencyKey).
   • Redact secrets, tokens, and sensitive fields from step logs and UI inspection payloads before persisting or rendering.
2. Honest Visual Run Overlay on the Graph:
   • When inspecting a historical or live run on the canvas, render the exact WorkflowVersion graph that the run executed, never the current modified draft.
   • Visually and textually distinguish every step state (using icons/badges + labels, never color alone):
     • succeeded, running, waiting (showing what it is waiting for: timer deadline, webhook correlation key, or approver), failed (showing attempt count and error), skipped (unchosen conditional path), cancelling / cancelled, and uncertain_side_effect (when an external call timed out after dispatch and requires reconciliation before blind retry).

Explicit Cancellation Semantics

• Distinguish Cancellation Requested (cancelling) from Cancelled (cancelled):
  • Cancel pending timers and prevent unstarted downstream steps from launching.
  • Clearly show steps whose external side effects already completed before cancellation arrived so operators never assume cancellation automatically undid an external payment or email.

End-to-End Workflow Verification Discipline

Verify workflow definitions and engines across the full matrix of operational failure modes:

• Branching & Join Matrix: Test true vs. false vs. default branches, skipped-branch propagation, and parallel join synchronization.
• Durable Wait & Process Restart: Trigger a run that enters a timer or approval wait, restart the worker/runtime, deliver the wakeup event, and assert the run resumes from its persisted checkpoint without re-running completed prior steps.
• Duplicate Event & Retry Idempotency: Deliver the same trigger webhook, approval decision, or step completion twice and assert zero duplicate step executions or duplicate external side effects.

Checklist

• Does the run inspector render the pinned WorkflowVersion graph and show runId, stepExecutionId, and per-attempt history?
• Are waiting, skipped, failed, cancelled, and uncertain_side_effect states clearly differentiated with non-color indicators?
• Are secrets masked in all recorded step inputs, outputs, and error traces?
• Have process restarts during waits, duplicate events, and step retries been verified end to end?