Test the step with realistic inputs
A single-step run is useful for checking an HTTP request, model action, or other supported step before you connect it to a larger chain. You can provide input data, outputs from earlier steps, or a source run for context. The returned result includes the step status, output, credits, timing, and any error or warnings.
A human approval step is not executable as an isolated test, and the step must exist in the workflow definition. A test call also does not prove that the complete workflow's branches, triggers, or downstream mappings are correct. It checks the chosen step under the inputs you supplied.
Validate the full chain separately
Once the step behaves as expected, run the workflow with a representative input and inspect each step's result. Check that templates reference valid upstream outputs and that the chosen trigger supplies the fields the workflow expects. A successful isolated step can still fail when its real upstream data differs.
Use test inputs that do not cause unwanted external effects. A step test may call a real service, send a request, or perform other work depending on the step type; it is not automatically a dry run. Confirm the step's effects before testing against production data.