Delegation

Parallel execution of AI agents is managed by applying decision rules that select the right delegation pattern based on task complexity, communication, and context overhead. This optimizes agent orchestration and prevents overcoordination by matching the delegation approach to specific workload requirements.

Task delegation uses lightweight one-shot workers for simple tasks and full agents for complex workloads requiring tool iteration. Decision rules evaluate communication, context, coordination, duration, and overhead to determine the appropriate effort-level scaling.

The best way to create an agent team for a complex project is by executing TeamCreate to spawn teammates, then coordinating tasks using SendMessage. This handles full agent workloads and enables parallel task dispatch across the team.

Ideal-state prompting improves AI agent briefs by defining the desired outcome for inline briefs and subagents, encouraging diverse perspectives and effective outcomes. This personalized approach ensures agents receive clear, targeted instructions for their specific tasks.

No specific dependencies are required to dispatch parallel tasks across isolated agents. The skill manages workloads across six patterns including built-in, isolated, background, custom inline agents, agent teams, and parallel task dispatch without external dependencies.