AI-Workflow-Learning-Lab

This paper documents the design, implementation, and lessons learned from AI-Workflow-Learning-Lab v2.0 — a constitutional governance engine for AI-native workflows. The project was built as a deliberate training exercise to explore and debug a practical development workflow combining VSCode, Cline (AI coding agent), Model Context Protocol (MCP), and Git. The implementation demonstrates core AMO research principles: formal authority boundaries, append-only ledger as a single source of historical operational truth, deterministic constitutional enforcement, constitutional self-validation, and replay-based integrity verification. A key finding is that AI-native development requires governance of the development process itself — not just of the runtime being built. The system enforces that execution requires prior legitimate authorization through a five-phase pipeline: submitted, reviewed, approved, validated, executed.