Skill Testing with Subagents

This skill implements a rigorous RED-GREEN-REFACTOR cycle specifically for process documentation and agent instructions. By simulating high-pressure scenarios—such as tight deadlines, sunk costs, and authority overrides—developers can identify exactly where an AI agent might naturally fail. The skill guides the iterative process of capturing these failures, drafting minimal rules to address them, and building rationalization tables to ensure the final skill is followed consistently, even in demanding real-world coding environments.

Key Features

Use Cases