What problem does it solve? Coordinating multiple AI agents on a software task is error-prone: agents self-certify their own work, skip reviews, and lose context between steps. This Skill provides a structured orchestration workflow that takes a GitHub Issue from initial research through planning, adversarial review, TDD implementation, and PR merge with enforced quality gates. ## Core Features & Use Cases - Multi-Agent Orchestration: Spawns specialized agents (Researcher, Architect, Coder, Code Reviewer, Security Auditor, PR Shepherd, Release Engineer) with defined roles and handoff protocols. - Gated Review Pipeline: Enforces a Plan Review Gate, a parallel Design Review Gate, and a 4-phase execution loop (Implement, Validate, Adversarial Review, Commit) where quality gates are blocking state transitions. - Task Tracking & Escalation: Manages work via markdown task documents in docs/tasks/, with human escalation after 3 failed iterations and GTG merge-readiness checks. - Use Case: A GitHub Issue labeled agent-ready is claimed automatically: the orchestrator researches the codebase, produces a reviewed implementation plan, decomposes it into work units, implements each with TDD and independent adversarial review, then opens and shepherds the PR to merge. ## Quick Start Start work on GitHub Issue 123 by saying: start task 123.
