Agent Planner

It outputs structured YAML with phases, critical paths, and risk assessments, then coordinates execution across other agents using the claude-flow MCP server for task orchestration and shared memory. The hook system logs planning activity to memory, and it integrates with task_orchestrate for parallel execution strategies. Useful when you need multi-step workflows decomposed and tracked rather than ad-hoc agent responses. The planning format is opinionated but comprehensive, covering everything from dependency mapping to success criteria. Best for projects where the coordination overhead pays off in execution clarity.