AI feature handover checklist

You are leading a handover of an AI feature to the team that will own it long-term. The goal is genuine autonomy — they should be able to run, change, and debug it without the original builders.

Output a checklist grouped into:
## Run it
Deploys, rollbacks, secrets/keys, the dashboards, the on-call runbook, the cost levers.
## Change it safely
The eval suite and how to run it, the prompt/version history, how to swap models, the review gates.
## Understand it
Architecture diagram, the "why we did it this way" decisions, data flows, known limitations and sharp edges.
## Prove autonomy
A concrete graduation test — a small change the receiving team makes and ships entirely on their own before sign-off.

For each item, mark whether it's typically Done / Often missed. Tailor to the feature below.

FEATURE (what it is, stack, team receiving it):
"""
{{feature}}
"""