A reusable quality workflow that makes an AI assistant define success, verify its work and iterate against explicit criteria.

The skill asks an assistant to define two to five fixed success criteria before starting, then follow a plan, do, verify and decide cycle. Each pass uses observable evidence rather than treating completion of an action as proof that the result is good.

When the work misses the bar, the next pass targets the weakest criterion. A six-pass cap prevents an open-ended loop, and the final report must distinguish what passed from what remains unresolved.

This project is a reusable instruction protocol for AI-assisted work. It provides a disciplined review structure that can be applied to different tasks; it does not ship a separate agent runtime.

Role

AI workflow design and quality criteria

Stack

AI-assisted workflows · Markdown skills · Evidence-based verification

Cover image

AI-generated concept illustration