What problem does it solve? Coordinating multiple AI agents on cross-cutting work risks lost commits, conflicting file edits, and premature shutdowns that destroy unintegrated work. This Skill provides an end-to-end workflow for running Claude Code Agent Teams safely, from team design through verified integration. ## Core Features & Use Cases - Team Design and Autonomy Control: Size teams of 3-5 teammates, choose Sonnet or Opus per task, and set self-organizing or lead-controlled autonomy based on task risk. - Worktree Isolation with Squash Contract: Isolate overlapping file edits in detached-HEAD git worktrees, requiring each teammate to squash work into a single commit before reporting. - Ordered Integration and Recovery: Cherry-pick commits in topological dependency order, resolve conflicts, run tests after every pick, and recover from crashed teammates or dirty worktrees. - Use Case: A feature spans frontend, backend, and database layers. Spawn three teammates in isolated worktrees, collect one commit hash per task, cherry-pick in dependency order, verify the full test suite, then shut down the team. ## Quick Start Ask the AI to coordinate this cross-cutting project with an agent team using worktree isolation and cherry-pick integration.
