team

Parallel multi-agent coordination is managed through a 5-stage pipeline: plan, prd, exec, verify, and fix. This approach decomposes tasks into independent stories, assigns agents to each stage, and ensures quality control through iterative fixes.

Task decomposition for parallel execution involves breaking down complex tasks into stories with clear dependencies and acceptance criteria. This allows independent stories to be assigned to agents and executed simultaneously while managing inter-story dependencies.

Verification in multi-agent execution checks completed work against pre-defined acceptance criteria. If issues are found, a bounded iteration loop triggers automated fixes until all criteria are met.

Task decomposition and parallel execution are ideal for large code refactoring across multiple modules. The pipeline parallelizes independent module updates, assigns agents to execute them, and iteratively fixes any integration issues found during verification.

Parallel multi-agent execution should be avoided for simple, linear tasks lacking independent components or dependencies. It is designed for complex scenarios requiring robust coordination; using it for trivial tasks adds unnecessary overhead.

Yes, the planning and prd stages require you to define clear dependencies and acceptance criteria for each story. This ensures agents execute their assigned tasks in the correct order and meet quality standards during verification.