Openspec

Spec-Driven Development, With Agents Doing the Typing

Why I hand agents a spec instead of a prompt, and how the explore → propose → apply → archive loop keeps AI output reviewable.