Vercel skills: install agent skills with npx skills, and what to check
How the npx skills command from Vercel Labs installs agent skills from GitHub into Claude Code, Cursor, Codex and more, where it writes and what to check.
- For
- Developers who found a skill on GitHub or skills.sh and want it installed
- Written for
- Claude CodeCodexCursorGitHub CopilotGemini CLIOpenCodeAny agent that reads SKILL.md folders
- Time
- 7 min read · updated Nov 21, 2026
The short answer
“Vercel skills” usually means npx skills, a command-line tool from Vercel Labs, MIT licensed, that installs agent skills from a GitHub repository (or another git URL) into the agent you use. Run npx skills add owner/repo from a project, pick the skills and agents, and it installs for that project; add -g to install for every project. Its README says it needs Node.js 22.20 or later and supports Claude Code, Codex, Cursor, OpenCode and 75 more. It sends anonymous usage data unless you set DISABLE_TELEMETRY=1, and it does not review what it installs, so list and read a skill first with --list.
Searches for “Vercel skills” land on a few different things: Vercel’s own skills for working with its platform, a skills directory at skills.sh, and the skills command that installs from either. This page covers the command, because that is the part you run. Everything about it below is from its own repository, read on 21 November 2026: the GitHub page, the README and the docs folder. It is a third-party tool and free to use, and we have not run it, since it reports usage, so these are its documented behaviors rather than our test results.
What it is
The repository, vercel-labs/skills, describes itself as “the open agent skills tool - npx skills”, with skills.sh as its home page. It is MIT licensed and its last push was on 17 November 2026. The README calls it the CLI for the open agent skills ecosystem and says it supports OpenCode, Claude Code, Codex, Cursor and 75 more agents. It requires Node.js 22.20 or later. You don’t install it first; npx runs it.
Install a skill
Step 1: List what a repository holds
Before you install, see what is in the repository.
--listprints the available skills without installing anything.TerminalStep 2: Install the one you want
Name the skill with
--skilland the agent with-a. Skills are installed into the current project by default. Add-gto install to your user folder so they are available across projects.TerminalStep 3: Restart the agent and check
Start a new session so the agent reads its skills folder again, then ask it which skills it has.
npx skills listshows what the tool installed.
Without --skill and -a, the command asks which skills and which agents. With -y it skips every confirmation prompt, which is fine for a repository you have read and a poor habit for one you haven’t.
What else it can do
The reference in its docs folder adds detail. A source can be GitHub shorthand (owner/repo), a full GitHub URL, a direct path to one skill in a repository, GitLab, Azure Repos, any git URL, a local path, or a direct download of a single SKILL.md or an archive. For private repositories it uses the authentication you already have configured, such as a git credential helper, GitHub CLI or SSH. --rename installs one skill under a new name, which matters because two sources can publish the same skill name and the second install overwrites the first.
Where it writes
The tool’s own table of agents and folders gives, for six common agents:
When you install interactively you can choose between a symlink, which the docs recommend, and a copy. A symlink points each agent at one canonical copy, so there is a single place to update; --copy makes independent copies for agents or systems where symlinks aren’t supported. In a project the tool also keeps a skills-lock.json that records each skill’s source and a content hash, so list can mark a skill as modified and check can report local changes and exit with code 1, which is useful in CI.
Check before you install
Installing a skill gives the repository’s author a way to put instructions, and sometimes scripts, in front of your agent. The tool installs what it finds; it does not vouch for it. A short routine before npx skills add:
Its documentation also says the tool collects anonymous usage data. GitHub repository and skill identifiers are sent only for repositories that GitHub confirms are public, other remote sources may include identifiers in install telemetry, and setting DISABLE_TELEMETRY=1 or DO_NOT_TRACK=1 turns it off. Safety checks before running a skill is the longer version of the list above.
How it relates to Lookbk
They do different jobs. npx skills installs from any repository or URL you give it. Lookbk’s installer installs Lookbk’s skills, each of which has a preview of what it rendered, with curl -fsSL https://lookbk.dev/cli.js | node - add <skill> and Node.js 20 or newer. The free ones are also in a public MIT repository, lookbkdev/lookbk-skills, whose README gives the npx skills line for it: npx skills add lookbkdev/lookbk-skills. That repository holds 38 skill folders in the skills/<name>/SKILL.md layout, the first one the tool’s discovery rules list. Its change log records that version 1.7.4 of the tool found the skills in a local copy, when the repository held seven; we have not repeated that check with the current version. The README adds that the CLI needs Node.js 22.20 or later and sends anonymous install counts to skills.sh, which DISABLE_TELEMETRY=1 turns off.
Whichever tool you use, restart the agent afterwards. Install a skill in any agent covers how to check it loaded in each one, and agent skills marketplaces shows how each agent adds a catalog of its own.
Common questions
- What is Vercel skills?
- Usually the
skillscommand from Vercel Labs, run asnpx skills, which installs agent skills from GitHub repositories and other git sources into coding agents such as Claude Code, Cursor, Codex and OpenCode. Its home page is skills.sh. - How do I install a skill with npx skills?
- Run
npx skills add owner/repofrom your project, choose the skills and agents, or name them with--skilland-a. It installs into the current project unless you add-g. - What does npx skills need?
- Its README says Node.js 22.20 or later.
npxdownloads and runs the tool, so there is nothing to install first. - Does npx skills check that a skill is safe?
- Nothing in its documentation says it does. Use
--listto see what a repository holds, then read eachSKILL.mdand its scripts before you install. - Does npx skills collect data?
- Its documentation says it collects anonymous usage data and sends repository and skill identifiers for repositories GitHub confirms are public. Set
DISABLE_TELEMETRY=1orDO_NOT_TRACK=1to turn it off. - Where does npx skills put a skill for Cursor?
- In
.agents/skills/inside the project, or~/.cursor/skills/with-g, according to its table of agents. Cursor’s own documentation says it loads skills from both folders. - Can I use npx skills to install Lookbk skills?
- The free ones, yes: the lookbk-skills repository’s README gives
npx skills add lookbkdev/lookbk-skills. Paid skills install with Lookbk’s own installer after you sign in.
Sources
What this guide says about other companies’ tools comes from their own documentation, checked on Nov 21, 2026.
Did this guide help?
Read next
All guidesInstall a skill in any agent
Install a Lookbk skill in Claude Code, Codex, Cursor, GitHub Copilot, Gemini CLI or OpenCode: where each looks, how to check it loaded and how to run it.
Where skills live: folders for every agent
The folders Claude Code, Codex, Cursor and other agents read skills from, which copy wins when two share a name, and what cloud agents can see.
Agent skills marketplaces: how each agent adds a catalog
There is no one skills marketplace for every agent. How Claude Code, Cursor, OpenCode, Gemini CLI, Codex and Copilot each add a catalog, from their own docs.
Claude skills on GitHub: main repos and how to use them
The GitHub repositories that hold Claude skills, checked on the GitHub API: license, last commit, skill count. Then how to copy one into your agent.
Safety checks before running a skill
A skill is instructions and code your agent may run. Six checks to make first: its source, its commands, its network use and what it may do without asking.
See a skill’s result before you install it.
Every skill page shows what it makes, what it needs and which agents can run it.



