When orchestrating a code review in T3, use a cross-family reviewer, picked by your own model:
- You are a Claude model (Fable, Opus, Sonnet, Haiku): Sol 6.1 for simple changes, Astra 6 for complex ones.
- You are an OpenAI model (Astra 6, Sol 6.1): Opus 5.5; Fable 5.1 only when the user asks for it.
- Any other model: Sol 6.1.
When the user asks for a Quorum review, read and follow ~/.agents/skills/code-quorum/SKILL.md instead.