What problem does it solve? Coordinating multiple AI agents on a multi-task project without a written plan leads to overlapping work, unclear task boundaries, and improvised branching. This Skill runs a structured planning phase on the agents-workbench branch that produces a scored, wave-organized plan before any agents are spawned. ## Core Features & Use Cases - Work decomposition with scoring: Breaks a project into independent tasks, scores complexity (1-4) and risk (likelihood x impact), and extracts shared infrastructure as a Task 0. - Wave planning and branch strategy: Groups more than 3 tasks into waves of at most 3, and requires an explicit branching strategy decision (per-task, shared feature branch, or single branch). - Plan artifact generation: Writes the plan to .agents/plans/<project-name>.md with a Task List, Risk Register, Dependencies Map, Path Ownership, and Success Criteria, then updates AGENTS.md. - Use Case: Before building rate limiting and audit logging into an API gateway, run the planning phase to get three approach options, scored tasks, a risk register, and a written plan that team-execute can pick up in a later session. ## Quick Start Ask the agent to run /team-plan with a project description such as "add rate limiting and audit logging to the API gateway" while on the agents-workbench branch.
