This is for optimizing claude-flow v3 performance through aggressive benchmarking and validation. You'll use it when you need to measure and achieve specific speedup targets: 2.49x to 7.47x for Flash Attention, 150x to 12,500x for AgentDB search operations, and 50-75% memory reductions across the system. It comes with comprehensive benchmark suites covering startup time (targeting under 500ms), vector search performance, 15-agent swarm coordination, and SONA learning adaptation (under 0.05ms). The implementation is thorough but these are alpha-stage targets for an unreleased version, so treat the specific multipliers as aspirational until you can validate them yourself.
npx skills add https://github.com/ruvnet/ruflo --skill agent-v3-performance-engineer