• Fix the onboarding flow so model selection, manual key entry, and cancellation all resolve correctly instead of re-prompting on every launch.
• Add a Prime Inference onboarding splash for Prime CLI users and persist an onboarding-completed flag.
• Make the installer ask before preparing the IPython kernel runtime, and avoid stalling on the uv prompt during setup.
• Reframe the system prompt as a general-purpose agent that uses code, with clearer guidance on shell vs. kernel state persistence.
• Drop grok-code-fast-1 from the Prime Inference model list.