Install a TrustedSkills Skill in Claude Code
Install a SKILL.md skill from TrustedSkills into Claude Code with one skills CLI command. Covers project vs global scope, where the files land, how to confirm it loaded, and how to remove it.
Last updated 27 September 2026
Run the install command from the skill's page in your project folder, adding -a claude-code. The skill lands in .claude/skills/<name>/ and Claude Code picks it up in your next session. No JSON to edit.
Almost every listing on TrustedSkills is a SKILL.md skill: a folder with an instruction file in it. It is not an MCP server, so there is no config file to edit and nothing to restart. One command copies it into place.
The Command
Every skill page shows its own install command. Copy that command and add -a claude-code. Using find-skills as the example:
npx skills add vercel-labs/skills --skill find-skills -a claude-code
That is the exact line the skill page gives you when Claude Code is your selected platform — this page and that box are generated by the same function, so they cannot disagree.
What You'll Need
- Node.js — the skills CLI runs through
npx. Check withnode --version; install from nodejs.org if it is missing. - Claude Code — see the Claude Code setup guide if you have not installed it yet.
This Project, or Every Project
Run the command with no flags and the skill is installed for the current project only. Add -g and it goes to your home folder, where every project sees it:
| Command | Where the skill lands | Who sees it |
|---|---|---|
npx skills add vercel-labs/skills --skill find-skills -a claude-code | .claude/skills/<name>/ in the current folder | This project |
npx skills add vercel-labs/skills --skill find-skills -a claude-code -g | ~/.claude/skills/<name>/ | Every project on this machine |
Project scope also writes a skills-lock.json next to the folder, recording each skill's source repo and a hash of its contents. Commit it and a teammate can restore the same version with npx skills experimental_install. One wrinkle worth knowing: the lock file does not record which agent you installed for, so the restore lands in the shared .agents/skills/ folder. To get it into .claude/skills/ specifically, re-run the add command with -a claude-code.
Confirm It Worked
Ask the CLI what is installed:
npx skills list
It prints each skill, the full path it was written to, and which agents it was installed for. If the skill you just added is not listed, you ran the command in a different folder than you think.
Then start a new Claude Code session. Skills are matched by their description field, so you do not call one by name — ask for the thing it does, and the agent reaches for it.
Updating and Removing
# Update every installed skill
npx skills update
# Remove one skill
npx skills remove find-skills
Leave -a off: with the flag, the CLI only clears that agent's links, which can report success while leaving the files in place. Without it, the folder goes. Nothing else on your machine is touched.
Troubleshooting
"command not found: npx"
Node.js is not installed, or your terminal was open before you installed it. Close the terminal, open a new one, and run node --version again. See what npx actually does.
The skill installed but the agent ignores it
Two usual causes. You installed into a project folder and are running Claude Code somewhere else — check the path in npx skills list. Or the session was already open: skills are read when the session starts, so start a new one.
The repo has several skills and I got the wrong one
Add -l to list what a repo contains without installing anything:
npx skills add vercel-labs/skills -l
Frequently Asked Questions
Do I need to edit a JSON config file to install a skill?
No. SKILL.md skills are folders on disk, not MCP servers. The claude mcp add command and the .mcp.json file are for MCP servers only.
Where does Claude Code keep installed skills?
.claude/skills/<name>/ inside the project for a project install, and ~/.claude/skills/<name>/ for a global one with -g.
Can I use the same skill in another agent?
Yes. Run the same command with a different -a target: cursor, openclaw, codex and opencode are all valid targets, and you can repeat the flag to cover several at once — -a cursor -a claude-code. A comma-separated list is rejected. See installing a skill in Cursor.
How do I know what a skill will do before I run it?
Open the SKILL.md the CLI wrote, or read it on the skill's page before installing. A badge on TrustedSkills never means a person reviewed the code — see what each badge actually claims.
TrustedSkills Team
The TrustedSkills team maintains the TrustedSkills index of AI agent skills. The index records where each skill comes from and how to install it. It does not review or audit skill code.