Start here
Start with a reviewed workflow handoff. Choose fictional inputs, an isolated destination, expected outputs, and the actions authorized for the test. Keep the first run in draft or preview form when possible.Copy this prompt
Use this in Screenpipe chat or an assistant with the required connections. Replace the brackets with your details. If the assistant lacks access, provide a reviewed excerpt; a prompt alone does not connect it to your history.Check the result
Inspect the actual destination and compare fields with expected results. Verify that failures leave the external system in an acceptable state. A run log saying “done” is insufficient.See a worked example
See a worked example
Go deeper: setup choices and edge cases
Go deeper: setup choices and edge cases
Use a small set of representative examples with an owner who knows the workflow. Keep failures and corrections in an evaluation log so a prompt change does not silently reintroduce an earlier error.Only add a schedule after manual tests pass. Define who responds to failures, retry limits, usage review, and how to disable future runs. Re-test when a provider, prompt, input format, or external system changes.