team-builder

To assemble AI agents for parallel execution, discover available agents from project-local and global directories, select them via an interactive domain menu, and spawn up to five agents simultaneously to synthesize a unified result.

Domain-based agent discovery scans project directories for markdown persona prompts, extracting the first heading as the agent name and first paragraph as the description, then organizing available agents into categorized domain menus for selection.

Format markdown files as agent personas by ensuring the first heading serves as the agent name and the first paragraph provides the description, adhering to the discovery rules required for domain-based organization and team assembly.

You can select specialized agents by both numbered and name-based selection across different domains, enabling you to compose a cross-domain team like security, performance, and data to compare perspectives side-by-side for complex tasks.

The maximum number of agents you can spawn in a single team is five. This limit ensures manageable parallel execution and result synthesis when coordinating specialized agents across different domains for complex tasks.