jasondasuki/sre-ai-skills ? reverse-engineered prompt

Build me a small repo for reusable personal AI agent skills, where each skill lives as a portable core with a SKILL.md, optional longer reference docs, and small helper scripts.

I want the core files to stay machine agnostic, so don’t hardcode any local paths, model names, MCP server names, or secrets in the shared part. Instead, make it easy to keep a thin per computer handler somewhere else that plugs in the working directory, runtime settings, and secret references.

Please include a simple onboarding guide for setting up a new machine, plus a helper script that can check whether a skill core and its handler match cleanly. Also add an easy way to create and maintain new skills from a template so I can add more over time.

Keep it lightweight and practical, focused on shell based helpers and plain markdown docs, and make sure the structure is clear enough that an agent could load the skill core on demand. If you need current best practices, look them up online.