This implements Goal-Oriented Action Planning (GOAP) for multi-step objectives. You define a goal state, inventory available actions with their preconditions and effects, and it generates an optimal plan using A* pathfinding through the state space. The interesting bit is adaptive replanning: it verifies preconditions before each action and recalculates the path if something fails or changes. It tracks everything through trajectory hooks and stores successful plans to memory. Use this when your task has real dependencies and you need more than a simple checklist. It's proper planning automation, not just task breakdown. The tradeoff is upfront modeling cost, but for complex deployments or migrations where order actually matters, that pays off.
npx skills add https://github.com/ruvnet/ruflo --skill goal-plan