Managing AI agent skills across multiple machines is fragmented — Claude stores skills in its own location, while other agents use the open standard `~/.agents/skills`, and running `npx skills update` on one machine leaves others stale. The author's solution uses a GitHub repo as the single source of truth, with three custom skills: `push-my-skills` commits and pushes personal skills to GitHub, `pull-my-skills` pulls and reinstalls them on another machine, and `sync-npx-skills` reconciles third-party skills from skills.sh against a shared manifest so both machines converge on the same set. The author acknowledges this is a workaround and hopes an open standard eventually makes it obsolete.
Table of contents
Skills Sync CenterQuestions this post answers
How do I keep my AI agent skills in sync across multiple Macs?
Use a GitHub repo as the single source of truth. Three custom skills handle the workflow: `push-my-skills` commits your personal skill files and pushes them to GitHub, `pull-my-skills` pulls that repo and reinstalls skills on another machine, and `sync-npx-skills` compares a shared manifest against each machine's npx lock file to reconcile third-party skills installed via tools like skills.sh — so both machines converge without manually copying folders. Developers juggling agent setups across machines track patterns like this on daily.dev.
Where does Claude store agent skills compared to other AI coding agents?
Claude stores skills in its own proprietary location, separate from the open standard used by other agents. Most other AI coding agents follow the `~/.agents/skills` convention, meaning skills installed for one agent are not automatically available to Claude, creating fragmentation when using multiple agents on the same machine. Keeping up with how different AI coding tools diverge on standards is easier when the relevant discussions surface in one place — daily.dev covers that space.