SkillCatalog

Claude Code, Cursor, and Codex can follow skills: Markdown instructions that each AI tool reads from its own skills folder on your machine. When a team shares skills by copying them from one machine to the next, each person soon has a different version, and nobody can tell which one is current.

SkillCatalog keeps a team's skills in Git repositories called catalogs, and writes the skills each person chooses into the folders their AI tools read. You choose the skills you want in every repository, and a repository can list skills of its own, so everyone who works in it gets the same ones. When someone changes a skill and syncs, everyone else gets the change on their next sync, after any review the catalog requires.

SkillCatalog runs on your machine, so there is no server to run and no account to create. It reaches your catalogs through Git with the access you already have, which means your Git host decides who can read or change each catalog.

SkillCatalog comes in two parts. The desktop app, for Macs, is where you browse, write, and sync skills. The command-line tool, skc, does the same work in a terminal, a script, or a CI job. The desktop app includes skc, and on Linux you install skc on its own.

Where to start

1. Install SkillCatalog.
2. Follow Getting started, where Acme sets up a catalog, delivers a skill to its AI tools, and shares one repository's skills with a teammate.
3. Read How SkillCatalog works for the model behind those steps and for what SkillCatalog leaves to you.

After that, each guide covers one task, and the reference pages, starting with the CLI reference, list every command, file format, and folder. If an AI agent will run skc for you, point it to SkillCatalog for AI agents.

Get help

When a command fails, look up its message in Troubleshooting, which gives the cause and the fix. skc <command> --help explains any command and its flags.

To report a problem or ask a question, email hi@skillcatalog.dev with the details that Report a problem lists.