Skill Autoresearch Loop

Make your skills better with a real eval loop instead of random prompt edits.

Skill Autoresearch Loop is for builders shipping reusable skills, prompt systems, or agent workflows that keep drifting over time. It gives you a measurable keep-or-revert process so improvements are based on evidence instead of vibes.

What you get:

• a baseline-and-compare loop for skill revisions
• structured testing against failure cases and weak outputs
• before-and-after evidence for deciding whether a change actually helped

Week-one proof looks like one skill revision evaluated clearly enough to keep the right change and reject the wrong one.