team-builder

The team-builder skill is designed to streamline the process of assembling specialized teams from a pool of AI agents, each defined by distinct personas. It addresses the challenge of managing multiple agents across different domains, allowing users to browse, select, and compose parallel teams efficiently for a given task. By providing an interactive menu, it simplifies the discovery and selection of agents, ensuring that teams are tailored to specific project needs without manual tracking of individual agent capabilities or domains.

This skill offers several key features. It supports both flat and subdirectory layouts for agent organization, automatically inferring domains either from folder structures or filename prefixes. Agents are sourced from multiple origins, including the CLI via the `claude agents` command, local project files, and global user files, with a clear precedence order to resolve conflicts. The system extracts agent names and descriptions from markdown files, and it dynamically generates domain-based menus for easy selection. Plugin agents, user agents, and built-in agents are all accommodated, providing a unified interface to manage diverse agent collections.

Team-builder is ideal for users who work with multiple agent personas and require ad-hoc team composition across various domains such as engineering, marketing, or sales. It is particularly useful for those needing to explore available agents before task assignment, for creating domain-specific or cross-functional teams, and for rapidly deploying specialized agents without manual configuration. Target users include developers, project managers, and anyone leveraging AI agents in collaborative or multi-domain environments.