swarm-planning-and-assignment

Parallel execution in swarm topology planning is managed by structuring tasks into parallelizable lanes with explicit dependencies between them. You identify work units, map their relationships, and assign them across linear, tree, or swarm topologies to enable concurrent agent execution.

To set up agent task assignments, you list work units from your spec, identify dependencies between them, assign explicit owners, and attach required skills, tools, models, and verification steps for each task to ensure complete traceability.

Swarm planning supports linear, tree, and swarm topologies for multi-agent task assignments. These topologies allow you to structure work with varying levels of parallel execution and dependency complexity depending on your workflow requirements.

You define required skills and tool policies for delegated agent tasks by attaching specific skill loadouts, model policies, and tool permissions to each task within the assignment structure. This ensures agents receive the correct capabilities and constraints for their assigned work.