What problem does it solve? When multiple repositories each run their own Squad (AI team), there is no shared runtime, MCP tools, or issue tracker between them. This Skill provides concrete protocols for exchanging knowledge queries, PR reviews, task delegations, and dependency analysis across those isolated squads. ## Core Features & Use Cases - Four communication patterns: synchronous Copilot CLI sessions (Pattern 0), read-only metadata queries (Pattern 1), git-based async requests (Pattern 2), and GitHub issue-based delegation (Pattern 3), with a decision tree for choosing between them. - Liveness monitoring: log-directory-based stall detection and recovery actions for long-running CLI sessions with slow MCP server initialization. - Structured request/response formats: YAML schemas for cross-squad requests and responses, plus a platform compatibility matrix covering GitHub Issues, Azure DevOps, and Planner. - Use Case: A research squad needs the platform squad's architecture decisions. It reads the peer's .squad/decisions.md directly for a quick answer, or spawns copilot -C <repo> --agent squad for deeper analysis, or files a labeled GitHub issue when the repo is not cloned locally. ## Quick Start Ask the agent to query a peer squad's architecture by reading its .squad/team.md and decisions.md files, or by spawning a Copilot CLI session with --agent squad against the target repository.