ls-spawnteam

Multi-agent orchestration for code review is handled by spawning a Claude agent team using the code-review template. This creates a structured workflow that partitions tasks, assigns owners, and runs parallel teammate calls to execute the review collaboratively rather than relying on a single agent.

Task decomposition for complex feature development is achieved by spawning an agent team with the feature development template. The system creates all sub-tasks before spawning teammates, assigns specific owners, and optionally uses plan-mode or dry-run flags to validate the decomposition strategy before execution.

A security audit can be executed by triggering the built-in security-audit agent team template. This approach orchestrates parallel Claude agents that follow a structured task pipeline, ensuring the audit is partitioned and assigned to specific teammates for comprehensive coverage.

Multi-agent orchestration supports both dry-run and plan mode flags during task decomposition. Using the dry-run path presents the planned decomposition and owner assignments for review before actually spawning the parallel teammate Task calls, ensuring the workflow is validated first.

Setting up a thinker debate or PRD pipeline involves triggering the respective built-in templates to spawn an agent team. The system parses the template and flags, creates the necessary tasks, assigns owners, and runs parallel teammate calls while maintaining team_name consistency throughout the workflow.