Phantom Builds.

Audits a named AI model for safety, reliability, capability, and related axes using cited public evidence, then returns a bias line, raw audit, and Grok Build team packet — then stops.