Workflow Orchestration

This is a structured approach to handling non-trivial tasks with planning docs, subagent delegation, and a self-improvement loop that tracks mistakes in a lessons file. You enter plan mode for anything with 3+ steps, offload research to subagents to keep your main context clean, and never mark work complete without verification. The honest take: it's heavy process for simple tasks but makes sense if you're building features or debugging complex issues where quality gates matter. The lessons tracking after corrections is the most interesting piece, it creates a feedback loop that actually sticks across sessions instead of repeating the same mistakes.