# The Registry
A skill registry for private-first distribution of skills, agents, prompts, output styles, and config files across devices, projects, and teams.
## What It Is
The Registry is a single skill whose only job is to manage other skills. It's a catalog of references — local file paths and GitHub repo URLs — that point to where your skills, agents, prompts, output styles, and config files live. Nothing is copied or installed until you ask for it.
Think of it as a `package.json` for agent capabilities — but instead of packages, you're managing skills, agents, prompts, output styles, and configs. Instead of a registry like npm, you're pointing at your own GitHub repos and local paths.
**This is a pure agent application.** No scripts, no CLIs, no dependencies, no build tools. The entire application is encoded in `SKILL.md` and a set of reference docs that teach the agent what to do. The agent IS the runtime.
## Why It Exists
As you build with AI agents, you accumulate skills, custom agents, and prompts across multiple repos and machines. You need to:
- **Reuse** them across projects without copy-pasting
- **Distribute** them to agents running on other devices
- **Share** them with your team without making everything public
- **Keep them private** — these are specialized capabilities
- **Stay in sync** — one source of truth, not stale copies everywhere
## How It Works
### The Catalog (`registry.yaml`)
```yaml
default_dirs:
skills:
- default: .claude/skills/
- global: ~/.claude/skills/
agents:
- default: .claude/agents/
- global: ~/.claude/agents/
prompts:
- default: .claude/commands/
- global: ~/.claude/commands/
output-styles:
- default: .claude/output-styles/
- global: ~/.claude/output-styles/
configs:
- default: .claude/
- global: ~/.claude/
registry:
skills:
- name: my-skill
description: What this skill does
source: /Users/me/projects/tools/skills/my-skill/SKILL.md
requires: [agent:helper-agent]
- name: remote-skill
description: A skill from a private repo
source: https://github.com/myorg/private-skills/blob/main/skills/remote-skill/SKILL.md
agents: []
prompts: []
output-styles:
- name: my-style
description: A custom output style
source: https://github.com/myorg/toolbox/blob/main/output-styles/my-style.md
settings:
outputStyle: "My Style"
configs:
- name: statusline
description: Status line script for the terminal footer
source: https://github.com/myorg/toolbox/blob/main/statusline/statusline.sh
settings:
statusLine:
type: "command"
command: "bash ~/.claude/statusline.sh"
```
The catalog stores pointers, not copies. Skills live in their source repos. You pull on demand.
**Output styles and configs** are single files: an output style installs as `<name>.md` under `output-styles/`; a config installs under its own basename (a `statusline.sh` stays `statusline.sh`). The optional **`settings:`** field holds the `settings.json` keys that activate the asset — `use` deep-merges them into the target scope's `settings.json` (entry keys win, everything else preserved, idempotent). Keep `settings:` values machine-portable: `~`-based paths, never `/Users/<name>/...`.
### Source Formats
| Format | Example |
| ------------------ | ------------------------------------------------------------------ |
| Local filesystem | `/absolute/path/to/SKILL.md` |
| GitHub browser URL | `https://github.com/org/repo/blob/main/path/to/SKILL.md` |
| GitHub raw URL | `https://raw.githubusercontent.com/org/repo/main/path/to/SKILL.md` |
The source points to a specific file. For skills the system pulls the entire parent directory (skills include scripts, references, assets — not just the markdown file); agents, prompts, output styles, and configs pull just the file.
For private repos, authentication uses SSH keys or `GITHUB_TOKEN` automatically.
### Typed Dependencies
Dependencies use typed references to avoid name collisions:
```yaml
requires: [skill:base-utils, agent:reviewer, prompt:task-router, output-style:my-style, config:statusline]
```
Dependencies are resolved and pulled first, recursively.
## Prerequisites
- **Claude Code** (or a compatible agent harness that reads `.claude/skills/`)
- **git** — for cloning sources and syncing the catalog
- **gh** (optional) — GitHub CLI for private repo access. Install: `brew install gh`
- **GitHub SSH key or `GITHUB_TOKEN`** — for private repos (not needed if using `gh auth login`)
- **just** (optional) — for justfile shortcuts. Install: `brew install just`
## Installation
### 1. Clone to Global Skills Directory
Clone into `~/.claude/skills/registry`. This path makes `/registry` available as a global slash command in Claude Code.
```bash
mkdir -p ~/.claude/skills
git clone <your-repo-url> ~/.claude/skills/registry
```
### 2. Configure
Open `~/.claude/skills/registry/SKILL.md` and update the `## Variables` section with your repo URL.
```markdown
- **REGISTRY_REPO_URL**: `https://github.com/yourname/the-registry.git`
```
### 3. Verify
Start a new Claude Code session anywhere. `/registry list` should work and show an empty catalog.
## Quick Start
Typical workflow: **build → catalog → distribute → use**.
### Add a skill to the catalog
```
/registry add deploy skill from https://github.com/yourorg/infra-tools/blob/main/skills/deploy/SKILL.md
```
### Use it in another project
```
/registry use deploy
```
Want it globally available on this machine?
```
/registry use deploy install globally
```
### Push changes back
```
/registry push deploy
```
### Sync everything
```
/registry sync
```
## Commands
| Command | What It Does |
| ----------------------------- | ---------------------------------------------------------- |
| `/registry install` | First-time setup — clone, configure |
| `/registry add <details>` | Register a new entry in the catalog |
| `/registry use <name>` | Pull from source into local directory (install or refresh) |
| `/registry push <name>` | Push local changes back to the source |
| `/registry remove <name>` | Remove from catalog and optionally delete local copy |
| `/registry list` | Show full catalog with install status |
| `/registry sync` | Re-pull all installed items from source |
| `/registry search <keyword>` | Find entries by name or description |
### Justfile Shortcuts
```bash
just install # First-time setup
just list # List catalog
just use my-skill # Pull a skill
just push my-skill # Push changes back
just add "name: foo, description: bar, source: /path/to/SKILL.md"
just remove my-skill # Remove from catalog
just sync # Re-pull all installed items
just search "keyword"
```
> **Note:** Justfile recipes use `--dangerously-skip-permissions` because the agent needs filesystem and git access to clone, copy, and push.
## Architecture
```
~/.claude/skills/registry/ # The Registry skill (globally installed)
SKILL.md # Agent instructions — the brain
registry.yaml # Your catalog of references
references/ # Command recipes + shared reference material
install.md
add.md
use.md
push.md
remove.md
list.md
sync.md
search.md
source-formats.md # Source parsing, fetch/push workflows
justfile # CLI shorthand for all commands
README.md # This file
```
## Design Principles
- **Private-first**: Built for your specialized skills. Not a public marketplace.
- **Reference-based**: The catalog stores pointers, not copies. Skills live in their source repos.
- **Pure agent**: No scripts, no build tools. The SKILL.md teaches the agent everything.
- **Agent-agnostic**: Default target is `.claude/skills/` but supports any directory for any agent harness.
- **Catalog, not manifest**: Entries define what's available, not what's installed. Pull on demand.
## Credits
Based on [The Library](https://github.com/disler/the-library) by [IndyDevDan](https://github.com/disler).
Lire le README source