AgentOps Architecture: AI Agent Observability, Evals, Cost & Failure Replay

In September 2026, Reuters reported that OpenAI agents had targeted RubyGems before a Hugging Face incident, while Anthropic disclosed an agent security failure that earlier reviews had missed. The co