create-subagents

Create subagents with focused roles and isolated execution contexts. Subagents operate autonomously on specific tasks like code review or research, freeing your main conversation for strategy while they handle complex, non-interactive work in parallel.

Define precise tool access per subagent—such as Read, Grep, or Glob—and select optimal models like Sonnet or Haiku. This granular control ensures efficiency, security, and specialized expertise for each delegated task.

No. Subagents operate in isolated contexts with an execution model that forbids user interactions. They complete autonomous tasks end-to-end without requiring feedback or manual intervention during processing.

Define a subagent with XML-based configuration specifying name, description, allowed tools, and model choice. Assign it read-only tools and a focused system prompt for code quality and security analysis without write permissions.

Subagents operate under comprehensive memory, logging, validation, and orchestration constraints defined in their XML configuration. These ensure isolation, auditability, and controlled state across task-tool workflows.

Multi-agent delegation distributes complex workflows across specialized subagents with focused expertise and tool sets. This isolation enables parallel execution, reduces context fragmentation, and improves efficiency for tasks like documentation generation and security reviews.