I like to switch computers and operating systems, but I want my AI prompts, skills, and preferences to follow me from computer to computer. The way I do that is by:

  1. Storing prompts and skills in a “Prompts” folder. That folder is in the same location on each of my computers (e.g., ~/path/to/prompts).
  2. Using git to track, commit, and pull changes with a private GitHub repo as my remote repository.
  3. Symlinking that folder into each AI agent’s expected skills directory. For example: ln -s ~/path/to/prompts/skills ~/.claude/skills

That single symlink covers Claude Code (which reads ~/.claude/skills/ natively) and OpenCode (which reads it as a fallback). Other agents like Pi get their own symlink to the same source folder.

I similarly use a dotfiles folder and repo for configuration files for my applications.

This setup means the prompts repo is agent-agnostic, the per-agent step is just a symlink, and I can swap between harnesses while keeping my skills available.