Quickstart

This walks through the five minutes from a fresh shell to having a synced skill visible in every enabled harness.

1. Bootstrap a fresh repo

agents init --scaffold --profiles personal

This creates ~/.config/agents/ with the lockfile, .agents/ state dir, and an initial git commit. No remote yet — that’s optional.

2. Install a skill from skills.sh

agents skills add vercel-labs/agent-skills --skill deploy-to-vercel -y

Behind the scenes agents:

1. Fetches the skill folder from the GitHub repo.
2. Records a lockfile entry with the GitHub tree SHA so updates are detectable.
3. Symlinks each enabled harness’s skills path into the canonical copy inside the agents repo.
4. Auto-commits the lockfile change. (Push happens once you wire a remote.)

3. Wire a remote — auto-sync activates

agents remote add git@github.com:you/agents-config.git

remote add sets origin and pushes the current branch upstream in one step. From now on every agents skills add / update / remove pulls, commits, and pushes without you ever typing git.

4. On a second machine

agents init git@github.com:you/agents-config.git --profiles work
agents apply

init clones the config repo. apply reads the lockfile, refetches every remote skill into a cold cache, and creates the same symlinks. Every enabled harness now sees the same skill.

Next

• Project-scoped skills — install a skill into a specific project’s harness-local skills dirs instead of user-global.
• Profiles — gate skills by machine (work / personal / devbox).
• agents CLI reference — every flag.