What problem does it solve? Coordinating multi-phase AI development workflows (interview, planning, execution, review, QA) manually is error-prone and loses progress on interruption. This Skill sequences stages through a uniform PipelineStage interface with persisted state, so long-running workflows resume from the last incomplete stage instead of restarting. ## Core Features & Use Cases - Configurable Stage Sequencing: Compose pipelines from built-in stages (deep-interview, ralplan, ultragoal, code-review, ultraqa) or custom stages implementing the PipelineStage interface. - State Persistence and Resume: Pipeline state is written to .omx/state/pipeline-state.json via the ModeState system, enabling resume from the last incomplete stage and HUD phase rendering. - Consensus-Gated Planning: The ralplan stage requires durable Architect-then-Critic approval evidence before execution proceeds, blocking progression on missing or non-approving reviews. - Use Case: Run the default Autopilot pipeline to take a feature request from requirements clarification through consensus planning, durable execution, merge-readiness review, and adversarial QA in one orchestrated flow. ## Quick Start Ask the AI to run the pipeline skill with the default Autopilot stages to take your feature request from interview through QA.