What problem does it solve? Coordinating multiple AI agents on a single coding task often leads to conflicting changes, lost progress, and unverified results. This Skill defines a deterministic multi-agent execution contract that assigns one DAG owner, enforces phase gates, and validates every worker's output through strict receipts. ## Core Features & Use Cases - Single-Owner DAG Orchestration: Select exactly one execution owner (omp or orca) per run, with a persisted execution-owner receipt preventing competing task graphs. - Phase-Gated Pipeline: Run planning, test scaffolding, parallel implementation, validation, testing, review, and security audit phases with blocking gates between them. - Strict Worker Receipts: Every dispatched worker returns a five-field receipt (owned_paths, changed_files, verification, blockers, next_required_step) under a strict JSON schema. - Use Case: When running /auto go SPEC-AUTH-001 --team, the main session dispatches planner, executor, tester, reviewer, and security-auditor workers in bounded batches, verifies each receipt, and only marks the SPEC implemented after the completion gate passes. ## Quick Start Ask your agent to run the pipeline for a specification, for example: execute /auto go SPEC-AUTH-001 --team and coordinate the agents through the full phase pipeline.
