What problem does it solve? Coordinating multiple AI agents on a shared codebase risks conflicting writes, stale results, and lost work when agents fail. This Skill defines a Supervisor-based orchestration model that decomposes work into a dependency graph and integrates results safely. ## Core Features & Use Cases - Dependency-aware decomposition: Breaks GitHub Issues into a task graph, spawning only Ready or stack-ready nodes within WIP and resource limits. - Immutable contracts: Enforces spawn contracts (base snapshot, execution_generation, policies) and result contracts (commit/ref, validation results) so stale generations are never merged. - Stacked PR delivery: Supports stacked pull requests with predecessor snapshot pinning, Draft PR lifecycle rules, and revalidation when predecessors change. - Use Case: Split a cross-boundary feature into three GitHub Issues, spawn isolated workers per node, and have the Supervisor integrate validated results in dependency order while a cold reviewer checks each candidate snapshot. ## Quick Start Ask the agent to decompose this GitHub Issue into a dependency graph and orchestrate parallel workers with immutable snapshots and stacked PRs.