# Keep Starter Skill Entry Separated

When importing Starter/Essential v1 into Premium, the runtime and session boundary are merged but the Starter guidance remains a separate Starter-owned skill entry. The existing `skills/starter-openapi` identity is retargeted from `starter-full` to the merged Premium runtime after `cp9 up` and Starter substrate activation, while keeping Starter review gates, docs, artifacts, and failure classification separate from Premium OpenAPI guidance so Starter product semantics are not blurred into the Premium skill. The retargeted skill keeps separate logical base URL inputs for the `tidb-mgmt-portal` cluster lifecycle surface and the `serverless-service` data-operation surface even when both resolve through the Premium public api-gateway entrypoint. The first retargeted proof keeps the current Train A lifecycle-smoke boundary only; broader Starter/serverless operations stay follow-up trains after the merged lifecycle proof is green. Retargeted Starter Train A owns product-level lifecycle evidence only: create, poll, endpoint/mysql proof, product delete, and post-delete verification. It no longer runs `cp9 destroy`; session teardown belongs to the final merged Premium proof cleanup stage after selected train artifacts have been collected.