agent-coordination

Agent coordination lets you define execution patterns—parallel, sequential, swarm, hybrid, or iterative—to run multiple specialized agents and skills simultaneously or in dependency order. Use Skill() and Task() commands to invoke components, then apply coordination patterns with phase-based synthesis and validation gates to automate complex multi-step workflows.

Parallel coordination runs code review and test execution simultaneously, reducing total workflow time. Define dependencies between agents, set quality gates for validation, and use phase-based synthesis to combine results into a final output—ideal for orchestrating software development activities at scale.

Yes. Agent coordination handles task dependencies and optimization across software development activities including code review, testing, feature implementation, and architecture validation. Assign sequential or parallel patterns based on task requirements, and enforce validation gates between phases to ensure quality.

Swarm coordination is optimal when you need multiple agents to collaborate dynamically on the same problem, sharing results iteratively. Use it for complex exploration tasks; reserve sequential patterns for strict ordering and parallel for independent tasks that can run simultaneously without feedback loops.

Hybrid patterns combine parallel and sequential execution but increase complexity in managing state and dependencies across phases. Validate that all agents complete their gates before progressing to dependent phases; iterative refinement loops may require additional monitoring to prevent infinite cycles or missed convergence.

Quality gates are optional but recommended. Define validation checkpoints between coordination phases to ensure each agent's output meets requirements before downstream agents consume it. This prevents cascading failures and maintains workflow integrity across complex multi-agent orchestrations.