fred-performance-reviewer

Review FRED changes for runtime performance, concurrency, and async correctness as an independent, skeptical senior reviewer. Use when reviewing anything on the agent execution loop, LLM call sites, tool invocation, KPI/log emission, middleware, shared clients (HTTP/model/DB), caches, or any code that runs per-turn or per-request under concurrent load. Priority order: LLM call latency first, tool call latency…