Epistemic vigilance engine for brain-modeled agentic AI — source reliability, claim consistency, and belief coherence evaluation