Common workflow errors
Trace problems through input mapping, readiness checks, approvals and delivery.
Open the run and find the first node that failed, paused or was skipped unexpectedly. Compare its resolved input with the upstream output and check the status of the selected capability.
| What you see | What to check |
|---|---|
| The run does not start | Supply required run inputs and inspect the editor’s configuration or capability-readiness issues. |
| The node receives an empty or wrong value | Compare its configured path with the upstream node’s actual output. Check whether that upstream node ran. |
waiting_approval | Open the pending approval and inspect the node, role if assigned, and action input. |
| A branch is skipped | Check the incoming edge conditions and values used to evaluate them. |
| A connection, browser or model fails | Check the route and the specific resource it needs, such as a domain grant, knowledge base, or model supporting the chosen mode. |
| The run fails after an external action | Check the external record and any receipt before retrying. A capture or later node can fail after the action has already taken effect. |
| Delivery is missing or uses the wrong version | Verify the project context and selected source artifact ID. Open the project’s deliverable to confirm the preview. |
Reference diagnostics
emptymeans no reference was entered.invalid_formatmeans the editor could not parse the expression. Pick a run input or upstream field again.missing_sourcemeans the source node is unavailable to the reference picker. Check the node ID and incoming connection.missing_pathmeans the picker cannot match the field. Check the defined schema and the actual output rather than guessing a field name.
For example, a Data operation may put a value under $.nodes.node_clean.output.result.cleanRows. A field in a schema can still be absent in a specific run.