goap-agent

Task decomposition breaks down complex goals into structured subtasks with dependencies. GOAP planning applies a six-phase cycle—analyze, decompose, strategize, coordinate, execute, synthesize—to map dependencies, assign agents, and produce executable plans saved in the plans/ folder.

Multi-agent orchestration supports parallel, sequential, swarm, hybrid, and iterative execution patterns. Choose parallel for speed when tasks are independent, sequential for strict ordering, and hybrid or iterative patterns for complex workflows requiring adaptive coordination and dynamic replanning.

Quality gates are built-in checkpoints that validate outcomes at each phase. They enforce quality criteria across domains—storage, API, testing, documentation—and trigger plan adaptation when validation fails, ensuring robust cross-domain orchestration.

Use GOAP planning for tasks with ambiguous requirements, cross-domain dependencies, resource constraints, or quality-critical workflows. Linear automation works for simple sequential steps; GOAP handles optimization opportunities and structured decomposition where multiple valid execution paths exist.

The GOAP planning cycle assigns agents during the strategize and coordinate phases based on task analysis. You invoke the agent with a goal; it decomposes the task and generates execution assignments, so specialized agents are mapped dynamically rather than pre-configured upfront.

Coordination maps task dependencies and enforces execution ordering through the planning cycle. Quality gates validate intermediate outcomes, and adaptive replanning adjusts agent assignments if conflicts or failures occur, maintaining consistency across parallel execution streams.