moai-workflow-research

What problem does it solve?

Provides a structured, repeatable methodology to optimize moai-adk components (skills, agents, rules, config) via iterative binary evaluations, reducing ad-hoc experimentation and human drift.

Core Features & Use Cases

• Iterative binary evaluation experiments to compare configurations and component changes (skills, agents, rules, config).
• Built-in 5-layer safety architecture, risk controls, and worktree sandboxing to ensure safe experimentation.
• Data location conventions for eval suites, baselines, experiments, and observations to support reproducible research lifecycle.
• Eval-suite schema definitions and a lightweight dashboard workflow to track status, results, and progress.

Quick Start

Initialize a new moai-adk workflow research session and start the first evaluation cycle.