agentskills-usage

name	agentskills-usage
version	1.0.0
description	Teaches AI agents how to use the AgentSkills Registry CLI to search, pull, and push skill bundles
author	agentskills
tags	["meta","tutorial","cli","agent"]

AgentSkills Registry Usage

You are an AI agent that can interact with the AgentSkills Registry to discover, download, and publish reusable skill bundles. Use the agentskills CLI tool to communicate with the registry server.

Prerequisites

• The agentskills CLI must be installed and available in PATH.
• A registry server must be running (default: http://localhost:8000).
• To publish skills, you need an API token configured via agentskills login.

Available Commands

Search for Skills

Find skills by keyword or tag:

agentskills search <keyword>

Example:

agentskills search code-review

Output:

NAME                VERSION  DOWNLOADS  DESCRIPTION
code-review-agent   1.2.0    42         Automated PR code review skill
code-review-style   0.3.0    18         Style-focused code review

Pull (Download) a Skill

Download a skill bundle and extract it to the current directory:

# Pull the latest version
agentskills pull <skill-name>

# Pull a specific version
agentskills pull <skill-name>@<version>

Example:

agentskills pull code-review-agent

This creates a code-review-agent/ directory containing:

code-review-agent/
├── SKILL.md         # Read this to understand the skill's instructions
├── scripts/         # Executable scripts you can call
├── references/      # Reference documents for RAG / few-shot learning
└── assets/          # Static templates and resources

After pulling a skill, read its SKILL.md to understand what it does and how to use it.

Create a New Skill

Scaffold a new skill bundle:

agentskills init <skill-name>

Then edit the generated SKILL.md to define:

• YAML frontmatter: name, version, description, author, tags
• Markdown body: instructions that tell agents what this skill does and how to use it

Push (Publish) a Skill

Upload a skill bundle to the registry:

agentskills push <path-to-skill-directory>

Example:

agentskills push ./my-new-skill

The CLI will validate SKILL.md, pack the directory into a .tar.gz bundle, upload it to the server, and verify the checksum.

Configure API Endpoint

By default, the CLI connects to http://localhost:8000. Override with the --api-url flag:

agentskills search code-review --api-url https://registry.example.com

Save API Token

To publish skills, save your token first:

agentskills login

Workflow for Agents

When you need a capability you don't have:

1. Search the registry for relevant skills: agentskills search <keyword>
2. Pull the skill you need: agentskills pull <skill-name>
3. Read the skill's SKILL.md to learn its instructions
4. Apply the skill's instructions, scripts, and references to your current task

When you've built a useful capability worth sharing:

1. Init a new skill: agentskills init <skill-name>
2. Write clear instructions in SKILL.md, add scripts/references as needed
3. Push to the registry: agentskills push ./<skill-name>