The TEAM is an open source project that provides a portable agent skill for AI assistants such as Claude, Cursor, and Kiro. It functions as a personal AI advisory council composed of 25 named advisor personas, each with a defined specialty or "lane," along with a chief of staff persona named CATYOUSHA. The system is designed to prevent a single agreeable answer from passing as a decision by orchestrating structured input from multiple perspectives. The project is hosted on GitHub and distributed as a collection of files, with SKILL.md as the primary entry point that loads at session start, while supporting reference files such as roster.md, pillars.md, and triggers.md load on demand only when needed. The tool offers two primary modes of interaction. Users can call an advisor by name to receive a direct answer in character and within their specific area of expertise; if a query falls outside that lane, the advisor will hand off to another persona rather than bluffing. For more complex or high stakes decisions, users can fire a structured trigger: DECIDE for a fast call involving two to three voices, DANGER or CRASH for a full panel that engages multiple advisors in a roundtable discussion. The chief of staff persona, CATYOUSHA, convenes the appropriate advisors, facilitates real disagreement, and closes each session with a clear verdict or actionable next step. Named calls do not require loading the full roster file, making them a lighter and cheaper option by default. Typical use cases include product development decisions, project strategy, business trade offs, and personal decision making. The system is product agnostic and can be applied to any company, project, or situation. Users are advised to keep sessions focused and short, as longer conversations accumulate context and increase cost. For unrelated questions, starting a fresh conversation is recommended. The skill is designed to be portable, functioning across different AI tools and platforms that support the Agent Skills format. For use in Kiro or similar tools with manual inclusion, users must upload a zip of the repository root and explicitly reference the steering files to load them into context.