What problem does it solve? Coordinating multiple specialist AI agents on a complex feature manually is slow and error-prone. This Skill automates multi-agent execution by decomposing a request into tasks, spawning CLI subagents in parallel, coordinating them through shared memory files, and enforcing verification and QA review loops before accepting results. ## Core Features & Use Cases - Automated task decomposition and dispatch: Breaks a request into priority-tiered tasks, classifies them into domain tags, and spawns agents via native CLI paths (Claude Code, Codex, Gemini) or the oma agent:spawn fallback. - Memory-based coordination: Uses configurable MCP memory files (session, task board, per-agent progress and result files) with strict ownership rules to prevent write conflicts between concurrent agents. - Iterative review and recovery: Runs mechanical self-checks, oma verify, and QA cross-review loops with retry limits, clarification debt tracking, and escalation when loops are exhausted. - Use Case: Ask to implement a full-stack authentication feature; the orchestrator spawns backend, frontend, and QA agents in parallel, monitors their progress files, verifies each deliverable, and compiles a final summary. ## Quick Start Ask the assistant to orchestrate the implementation of your feature in parallel using the oma-orchestrator, providing the task description, workspace path, and acceptance criteria.
