Use this checklist before and during your first assigned Eval Labs workflow.
Before starting
- I know who assigned the work.
- I know the behavior family being tested.
- I know which surface I should use.
- I know not to use owner/admin surfaces.
- I understand that platform readiness is not human Lucia-quality approval.
- If I am a tester, I am using only Custom Prompt Test or Auto-generated Prompt Test.
While running
- I entered only the assigned prompts.
- I did not add unrelated cases mid-run.
- I waited for Lucia responses to complete.
- I opened only my own assigned/scoped run and Review Queue routes.
- I paused if the UI took me somewhere unexpected.
While reviewing
- I read the prompt and response before scoring.
- I treated suggested selections as suggestions, not truth.
- I scored honestly.
- I wrote short notes only when useful.
- I flagged uncertainty or risk for senior review.
- I saved every item.
Before finalizing
- Every item has been reviewed.
- Any uncertainty has been marked or escalated.
- I am finalizing my own assigned/scoped run.
- I am not using Team Review, Global Analysis, or owner/admin tools.