recipe-agent-team-architecture-design

Planning agent-team architecture upfront requires producing design artifacts such as architecture-candidates.md and technical-design.md to define module boundaries, task contracts, and interfaces before coding starts.

You compare architecture directions through candidate discovery, identifying options like A, B, and C, then capturing their tradeoffs in an architecture-candidates.md artifact to evaluate each design path.

You generate implementation tasks by converting the chosen architecture candidate into a coding-ready design, which outputs an implementation-tasks.json file containing bounded development actions with clear data contracts.

You need architecture design artifacts when planning module boundaries, interfaces, and task contracts in agent-team workflows to ensure aligned implementation and clarify downstream development scope.

Yes, architecture design for agent teams captures risks and questions in dedicated artifacts alongside technical-design.md and implementation-tasks.json outputs to ensure comprehensive upfront planning.

This approach applies to any agent-team workflow planning involving module boundaries, interfaces, task contracts, and candidate evaluation across design directions, regardless of the specific implementation platform.