Review Worker Runs
Understand run status, output previews, task links, and errors.
Worker runs show execution history for a worker.
Use run history before changing instructions. It shows whether the issue came from the prompt, the selected context, tool access, permissions, or the source record.
When to Review Runs
Review runs after activating a new worker, after changing instructions, after a failed job, and before expanding worker access. Run history is also useful when a teammate reports unexpected output but cannot remember which source record triggered it.
Run Status
Runs can appear as queued, pending, running, completed, or failed. The worker detail page refreshes while active runs are still pending, queued, or running.
Review a Run
Open a worker and review the Runs section. Each run can show:
- Status.
- Created time.
- Linked task when available.
- Task comment preview.
- Error message when the run failed.
Open the linked record whenever possible. The preview explains what happened, but the source record shows whether the worker had enough context and whether the final output is useful in the real workflow.
Review the run from source to output: trigger, source record, context available, tools allowed, worker instructions, result, and any linked task comment. This is the fastest way to tell whether the worker behaved incorrectly or the input was incomplete.
Task Links
If a run is linked to a task, use the task link to review the source task and the final worker output in context.
Check whether the worker output created duplicate comments, missed the requested format, or referenced data that was not present on the task.
Error Messages
If a run shows an error, review the worker instructions, context sources, tool categories, and permissions before running it again.
Common causes include missing record access, a paused or incomplete worker, instructions that require a disabled tool category, or source records that no longer exist.
Compare Runs
Compare a successful run with a failed or low-quality run. Look for differences in source record, status, context available at run time, tool calls, and output location.
Do not rerun repeatedly until you understand whether the previous run already created records, comments, or messages.
If a run created a customer-visible comment, task update, or message, review that output before retrying. Re-running may create duplicate updates unless the workflow is designed to replace or summarize previous output.
Decide the Next Action
Use the run review to choose the smallest useful change:
- Update instructions when the worker misunderstood the task.
- Add context when the worker lacked required information.
- Remove context or tools when the worker used information it should ignore.
- Pause the worker when failed runs may affect customers or shared records.
- Leave the worker unchanged when the source record was incomplete.
Quality Review Checklist
For low-quality output, check:
- whether the source record had enough detail
- whether the worker had the right context sources enabled
- whether instructions were too broad or conflicting
- whether the output format was specified clearly
- whether tool access was broader than needed
- whether the worker should be paused while changes are tested
Update one thing at a time when possible. If you change instructions, context, and tools together, it becomes harder to know which change improved the next run.
Troubleshooting
If the run list does not update, refresh the worker detail page after the active run leaves queued or running status.
If a run is marked completed but the output is missing, open the linked task or record and check permissions. The output may be attached somewhere you cannot currently view.