goal-plan

Define the goal state with concrete success criteria, assess the current state, then inventory available actions with their preconditions, effects, and costs. The Skill generates an optimal action sequence using A* search through the state space and creates tracked tasks for each step.

GOAP (Goal-Oriented Action Planning) is a planning technique where actions are selected based on preconditions and effects rather than fixed sequences. Use it when objectives have multiple dependent steps, changing conditions, or a need for adaptive replanning during execution.

When an action fails or produces unexpected results, the Skill reassesses the current state, recalculates the optimal path from the new state, and updates remaining tasks. Replanning triggers include failed preconditions, side effects, changed goals, exceeded cost thresholds, and unavailable dependencies.

Yes, it requires the ruflo-core MCP plugin tools for task creation, memory storage, workflow execution, and trajectory tracking hooks. It also uses standard Bash, Read, Write, and Edit tools for executing plan actions.

Avoid GOAP for simple linear tasks with no branching dependencies, where a fixed checklist is sufficient. The overhead of state assessment, cost estimation, and trajectory tracking only pays off when plans may need to change mid-execution.