michaelshimeles/skills ? reverse-engineered prompt

Build me a small Python based repo that packages a set of agent skills for Claude Code, Cursor, and Codex, plus an AGENTS.md workflow template that ties them together.

I want the repo to show a clear workflow for starting in an isolated worktree, organizing changes with a service layer style approach, proving the work with evidence, and then shipping with before and after proof plus a Greptile review loop. Include separate skill folders for things like visual before and after screenshots, service layer guidance, evidence driven testing, and the Greptile review loops, with each skill loading on demand from a SKILL.md file.

Make the README easy to understand for someone who just wants to copy these skills into their own project. Add a couple of simple tests for the evidence recorder piece if needed, and keep the whole thing practical and reusable. If any current docs or CLI behavior matter, look them up online so the instructions stay accurate.