What problem does it solve? Manually driving a multi-phase implementation plan through coding, code review, and fix cycles is slow and error-prone. This Skill automates the entire pipeline: it executes each phase, runs multi-agent code review, applies fixes, and repeats until the plan is complete, then archives results and compounds lessons learned. ## Core Features & Use Cases - Autonomous phase loop: Runs execute -> review -> fix per phase with a maximum of 2 fix cycles, tracking state via task checkboxes so it can resume after interruption. - Severity-gated decisions: Classifies review findings as P1/P2/P3, stops on unresolved blocking issues, and gracefully moves remaining P2 items into a known-issues file. - Completion pipeline: Performs final validation (typecheck, lint, tests, expo-doctor), archives the task, and documents solved problems into a knowledge base with learned-pattern rules. - Use Case: You have a planned refactor documented in docs/active/auth-refactor with phased task checklists. Invoke the autopilot on that folder and it implements every phase, reviews each with five parallel review agents, fixes findings, and delivers a final report with commits and known issues. ## Quick Start Run the autopilot on my task folder docs/active/auth-refactor and let it execute all remaining phases autonomously.
