agent-team-execution

To execute parallel tasks, you need a plan with discrete tasks at `docs/plans/*.md` and a reviewer for completion validation. The skill sets up a specialized agent team and orchestrates execution using shared contracts and SendMessage for coordination.

Coordination for parallel task execution uses SendMessage for all communication, avoiding file-based polling or locking. A single markdown shared contract file provides the data models and API signatures needed for synchronized task completion across specialists.

A shared contract is a single markdown file containing shared data models, API signatures, and environment config. You need it when orchestrating multi-agent teams to ensure all parallel specialists operate with consistent interfaces and coordinated execution.

Yes, this handles complex task management by facilitating parallel execution of 3+ independent tasks, each assigned to a specialist. A reviewer approves each task before final completion, ensuring quality and adherence to your specifications.

Limitations include the strict requirement for a pre-existing plan with discrete tasks at `docs/plans/*.md`. Additionally, the workflow mandates a reviewer to validate task completion, meaning tasks cannot finalize independently without this review step.

The best way is using a single markdown file as a shared contract for data models, API signatures, and environment config. This ensures all specialized agents access consistent definitions during parallel execution without file-based polling.