This spawns Agentica's multi-agent patterns after you've picked one through the orchestrator. You get four patterns: Swarm for parallel perspectives (like having security, performance, and architecture experts all weigh in), Hierarchical for coordinated implementation with planners and reviewers, Generator/Critic for iterative refinement over multiple rounds, and Jury for consensus validation with configurable voting. Each pattern sets environment variables so agents know their role and context. It's really just the execution layer, you're not meant to call it directly. The orchestrator handles pattern selection, this handles spinning up the actual agent configurations with the right premises and coordination logic.
npx skills add https://github.com/parcadei/continuous-claude-v3 --skill agentica-spawn