god-ai-architect

Design a multi-agent AI system architecture by defining agent roles, data flows, and risk mitigation steps. Use orchestrator/worker patterns, memory layers, and tooling strategies to ensure safe, scalable deployments.

ReAct and Plan-and-Execute are agent architectures for different reasoning needs. ReAct interleaves reasoning and actions dynamically, while Plan-and-Execute separates task planning from execution to improve reliability and control.

Secure AI systems against prompt injection by applying threat modeling, enforcing guardrails, and implementing validation checks. Add system-level security and observability through logging to monitor and block malicious inputs.

Evaluate AI architecture reliability and cost by applying cost-aware routing, observability metrics, and evaluation criteria. Analyze memory strategies and tool use to ensure the production deployment remains scalable and secure.

Yes, you can add guardrails and validation to existing multi-model AI deployments. The skill enforces security measures, observability, and system-level checks to enable safe, scalable production environments.