How to Use add-skill to Instantly Add Powerful Skills to Your AI Coding Agents

If you’re working with AI coding assistants like Cursor, Claude Code, OpenCode, Codex, GitHub Copilot, or any of the other 20+ supported tools in 2026, you’ve probably noticed something: they’re incredibly capable generalists, but they often need specific guidance to match your team’s conventions, project standards, or workflow preferences.

That’s where Agent Skills come in — a shared, open format that packages reusable instructions into neat, portable modules. Each skill lives in its own folder with one key file: SKILL.md. Inside that file you’ll find clear YAML frontmatter (name + description) followed by detailed Markdown instructions the agent follows when the skill is relevant.

The problem? Manually copying these skills into every tool’s folder quickly becomes tedious — especially if you use multiple agents or switch projects often.

Enter add-skill, an open-source CLI tool from Vercel Labs that solves exactly this pain point. With one command you can pull skills from GitHub (or any git host), have the tool automatically detect which agents you have installed, and place the skills in the correct directories — project-level or global.

In this guide we’ll walk through what Agent Skills actually are, how add-skill works, practical ways to use it, how to create your own skills, compatibility notes across tools, and answers to the questions developers ask most often.

What Are Agent Skills and Why Should You Care?

Think of Agent Skills as “prompt packages” you can install once and reuse everywhere.

Instead of repeatedly telling your AI things like:

  • “Always prefer server components in Next.js”
  • “Include Linear ticket links in every PR description”
  • “Follow our exact commit message and changelog format”

…you write those instructions once in a SKILL.md file. The agent then loads the skill automatically when the context matches (for example, when you ask it to “create a PR” or “write a new React component”).

Benefits you’ll notice quickly:

  • Consistency — the whole team uses the same high-quality guidance
  • Efficiency — no more copy-pasting long prompt blocks
  • Context savings — agents usually only load the skill name + one-line description until the skill is actually triggered, keeping token usage reasonable
  • Portability — one skill works across Cursor, Claude Code, OpenCode, and many others that follow the same specification

Real-world examples people use every day:

  • Generating release notes from git history
  • Enforcing company-specific PR templates
  • Automatically suggesting design-system components
  • Integrating with Linear, Notion, Jira, etc.

Introducing add-skill — Your Skill Manager CLI

The tool lives at https://github.com/vercel-labs/add-skill (install via npx — no global install required).

Basic command:

npx add-skill vercel-labs/agent-skills

What happens behind the scenes:

  1. Clones or downloads the repository
  2. Scans for valid skill folders (looks in skills/, root, .cursor/skills/, etc.)
  3. Detects which supported agents exist on your machine by checking known config directories
  4. Copies or symlinks the skills into each agent’s skill folder
  5. You restart your agent (or reload the window) → new skills are available

Supported Agents (as of early 2026)

Here’s a snapshot of the 23+ agents add-skill can target:

Agent Flag Project-level folder Global folder
Cursor cursor .cursor/skills/ ~/.cursor/skills/
Claude Code claude-code .claude/skills/ ~/.claude/skills/
OpenCode opencode .opencode/skills/ ~/.config/opencode/skills/
Codex codex .codex/skills/ ~/.codex/skills/
GitHub Copilot github-copilot .github/skills/ ~/.copilot/skills/
Gemini CLI gemini-cli .gemini/skills/ ~/.gemini/skills/
OpenHands openhands .openhands/skills/ ~/.openhands/skills/
Kiro CLI kiro-cli .kiro/skills/ ~/.kiro/skills/
… (and 15+ more)

Full list always available in the project README.

Practical Ways to Use add-skill

1. Install Vercel’s Curated Skill Collection (Best Starting Point)

npx add-skill vercel-labs/agent-skills

The CLI will usually ask which agents you want to target. Pick the ones you use.

Want to skip questions?

npx add-skill vercel-labs/agent-skills -y

2. Preview Skills Without Installing

npx add-skill vercel-labs/agent-skills --list

3. Install Only Specific Skills

npx add-skill vercel-labs/agent-skills --skill frontend-design --skill release-notes

4. Target One or More Agents Explicitly

npx add-skill vercel-labs/agent-skills -a cursor -a claude-code

5. Install Globally (Recommended for Personal Machines)

npx add-skill vercel-labs/agent-skills -g -y

Global skills live in your home directory and are available in every project.

6. Install from Any Git Repository or Subfolder

# GitHub shorthand
npx add-skill your-username/your-repo

# Specific subfolder
npx add-skill https://github.com/vercel-labs/agent-skills/tree/main/skills/pr-convention

# GitLab, Bitbucket, SSH, etc.
npx add-skill git@gitlab.com:team/skills.git

How to Create Your Own Skill (5-Minute Guide)

  1. Make a new folder (name it whatever you want)

  2. Create SKILL.md inside it

Example:

---
name: linear-pr-link
description: Automatically include Linear ticket link in PR descriptions and titles
---

# Enforce Linear Ticket Integration

## When this skill should activate

- User asks to create PR / pull request
- Branch name or commit message contains LINEAR-123 style ticket ID
- User mentions "Linear", "ticket", "issue"

## Required behavior

1. Extract Linear ticket ID from branch or latest commits
2. Fetch ticket title from Linear API if possible (or use placeholder)
3. Generate PR title: `[Feature/ Fix] Short description (LINEAR-123)`
4. Add to PR body:
   - Linear: https://linear.app/your-team/project/issue/LINEAR-123
   - Link to ticket discussion if available
  1. Push the folder (or the whole repo) to GitHub / GitLab

  2. Share the repo URL — anyone can now install it with one command

That’s it. The format is deliberately simple so teams can maintain their own internal skill libraries.

Compatibility Quick Reference

Not every agent supports every possible skill feature yet. Here’s a realistic 2026 overview:

Feature Cursor Claude Code OpenCode GitHub Copilot Kiro CLI Gemini CLI
Basic instructions Yes Yes Yes Yes Yes Yes
allowed-tools Yes Yes Yes Yes No Yes
context: fork No Yes No No No No
Hooks / events No Yes No No No No

Most teams stick to basic instruction-based skills — those work almost everywhere.

Frequently Asked Questions

I installed skills but my agent isn’t using them. What’s wrong?

Some agents (especially Kiro CLI) require you to manually add a resource glob to their config file. Check the agent’s docs for the exact step. Also confirm the SKILL.md has valid YAML frontmatter.

How do I remove a skill?

The CLI doesn’t have a remove command yet. Just delete the skill folder from the agent’s skills directory (project or global).

Can I install from a private repo?

Yes — use SSH URLs (git@github.com:...) or personal access tokens if the host requires authentication.

Does it work offline?

After the first clone, yes — but initial install needs git / network access.

How do I disable telemetry?

Set either environment variable:

DISABLE_TELEMETRY=1 npx add-skill ...
DO_NOT_TRACK=1    npx add-skill ...

CI environments disable it automatically.

Final Thought

In 2026, the difference between “good” and “great” AI-assisted coding often comes down to whether the agent has your team’s specific knowledge loaded. Agent Skills + add-skill make that knowledge portable, versionable, and dead simple to share.

Try it right now — it takes about 60 seconds:

npx add-skill vercel-labs/agent-skills -y

Open Cursor (or whichever agent you use), ask it to create a pull request following your company’s rules, and watch how much smarter it suddenly feels.

Have a favorite skill or a workflow you’ve automated? Drop it in the comments — I’d love to hear how you’re using this ecosystem. Happy coding!