Skills

skills is a command-line tool for installing and managing "agent skills," reusable SKILL.md instruction files that extend what coding agents can do. It's built for developers who use multiple AI coding agents (Claude Code, Cursor, Codex, OpenCode, and dozens more) and want a single, consistent way to add, update, and remove skills across all of them instead of manually copying files into each agent's config directory.

Skills are just markdown files with YAML frontmatter (a name and description) that tell an agent how to do a specific task, like generating release notes from git history, following a team's PR conventions, or integrating with tools like Linear or Notion. The CLI pulls these skills from GitHub, GitLab, other git hosts, direct URLs, or local paths, and installs them into the right directory for whichever agent you're targeting.

Key features

• Wide agent support: install skills into 79+ agents (Claude Code, Cursor, Codex, OpenCode, Windsurf, Gemini CLI, and more) using a shared --agent flag and per-agent path mapping.
• Flexible sources: add skills from GitHub shorthand (owner/repo), full GitHub or GitLab URLs, any git URL, local paths, or direct SKILL.md/archive download URLs.
• Private repo support: uses existing Git credential helpers, GitHub CLI auth, or SSH, so private skill repos work with the same commands as public ones.
• Project or global scope: install skills into a project directory (shared with your team via version control) or globally under your home directory (available across all projects).
• Symlink or copy installs: symlink skills to a canonical copy for easy updates, or copy independent files per agent when symlinks aren't supported.
• Use without installing: skills use generates a one-off prompt for a skill or launches a supported agent interactively, no permanent install needed.
• Lifecycle commands: list, find, update, remove, and init cover searching, listing installed skills, updating to latest versions, removing skills, and scaffolding a new SKILL.md.
• CI-friendly flags: -y/--yes and --all allow fully non-interactive installs for scripts and pipelines.

Ideal use cases

Use skills when you work with more than one AI coding agent and want to avoid re-implementing the same instruction sets (release notes, PR conventions, style guides) separately for each tool. It's a good fit for teams that want to version-control shared agent behavior inside a repo and have every contributor get the same skills automatically.

It also works well for individuals who want a quick way to try a skill without committing to an install: skills use generates a prompt or launches an agent interactively for a single, temporary use.

It is not a fit if you only use one agent that already has its own native skill/plugin system you're happy with, or if you need something beyond simple markdown-based instruction files (this tool doesn't run code or manage runtime plugins; it manages SKILL.md files and where they live on disk).

Installation

There's no separate install step; run the CLI directly with npx:

# Install a skill from a GitHub repo
npx skills add vercel-labs/agent-skills

Install a specific skill to a specific agent:

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

Use a skill once without installing it:

npx skills use vercel-labs/agent-skills --skill web-design-guidelines --agent claude-code

Other common commands:

npx skills list          # list installed skills
npx skills find react    # search for skills by keyword
npx skills update        # update installed skills
npx skills remove        # remove installed skills
npx skills init my-skill  # scaffold a new SKILL.md

For private repositories, use the same npx skills add commands; the CLI relies on your existing Git credential helper, GitHub CLI authentication, or SSH keys. You can also set GITHUB_TOKEN or GH_TOKEN explicitly for GitHub API access when needed.