Install an agent skill, or paste it into Grok Bot

MeloSkill listings already print an install command. Copy it. You do not need an account here. The command talks to GitHub and to the agent folders on your machine. Grok Bot does not use those folders — paste the file, in the section below. New to the file shape? Read what SKILL.md is first, then come back here.

Skills CLI

The usual command is from the open skills CLI (the same family as skills.sh). From a listing on this site it looks like one of these:

# Whole repo
npx skills add anthropics/skills

# One folder inside a repo (what most MeloSkill listings use)
npx skills add https://github.com/anthropics/skills/tree/HEAD/skills/frontend-design

# One named skill from a catalog
npx skills add vercel-labs/agent-skills --skill react-best-practices

Worked examples: frontend-design · frontend design task · taste-skill · taste task · shadcn · image to code.

Useful flags:

• -g — install for your user, not this git repo.
• -a claude-code / -a cursor / -a codex — only those agents.
• -s skill-name — one skill when the repo contains many.
• -y — skip prompts once you trust the source.

Prefer the command on the listing over inventing a path. Monorepos often have several SKILL.md files; installing the repo root can dump marketing or PDF skills you did not ask for.

Install / paste into Grok Bot

Grok Bot does not read .claude/skills or .cursor/skills. Open the listing, copy the SKILL.md (or the prompt on the task page), and paste it into the Grok Bot chat. Attach the file if that chat allows attachments. Then ask it to follow that file for the job.

For an invoice reminder, a newsletter, or posts on X, start from the Grok Bot playbooks instead of a frontend skill. Each playbook has an instruction you can paste, a weekly cap, and what to do when a plugin is missing.

Pasting text here does not install it into Grok Bot. MeloSkill is not affiliated with xAI or Grok Bot.

Claude Code

Claude Code looks for skills under .claude/skills/ in the project, or ~/.claude/skills/ when you pass -g. Each skill is a folder that contains SKILL.md (plus optional references/ or scripts/).

Typical install for one listing:

npx skills add https://github.com/anthropics/skills/tree/HEAD/skills/frontend-design -a claude-code

# Same skill for every repo on this machine
npx skills add https://github.com/anthropics/skills/tree/HEAD/skills/frontend-design -a claude-code -g

After install, open .claude/skills/ (or the global path) and confirm the folder name and SKILL.md are there. Restart the Claude Code session if the agent does not pick the file up. Then ask it to do the job — for example a blank-page UI with frontend-design, not “list my skills.”

No Node? Clone the GitHub path from the listing, keep that skill directory intact, and place it under .claude/skills/ yourself.

Cursor

Cursor reads project skills from .agents/skills/ or .cursor/skills/, and global skills from ~/.cursor/skills/. The CLI creates the path your Cursor version expects; if the listing’s command and disk disagree, trust the command, then confirm SKILL.md exists.

Typical install:

npx skills add https://github.com/leonxlnx/taste-skill/tree/HEAD/skills/taste-skill -a cursor

# User-wide
npx skills add https://github.com/leonxlnx/taste-skill/tree/HEAD/skills/taste-skill -a cursor -g

Commit the project skills folder when the team should share the same file. For a taste or landing job, start from the taste task; for a general blank UI, use the frontend design task. Read the SKILL.md on the listing before you stack a second overlapping file — two “make it look designed” skills will compete on triggers.

Codex

Codex uses .agents/skills/ in the project and ~/.codex/skills/ globally. Same npx skills add flow; pin the agent with -a codex when you do not want Claude or Cursor folders updated in the same run.

npx skills add https://github.com/anthropics/skills/tree/HEAD/skills/frontend-design -a codex

Frontend skills are ordinary instruction files here too — not a separate Codex marketplace. Pick the short list on a task page, install one path, then ask Codex to build against that brief.

Where the files go

Default is project scope. -g is global. Paths differ by agent; the CLI creates the right one. Agents and CLI versions move these paths — if a listing’s command and this table disagree, trust the command you copied.

Did it work?

Ask the agent to do the job the skill describes — “design this landing page” for frontend-design or taste-skill, not “list my skills.” If nothing happens, the YAML description is usually too vague for the trigger, or the file landed in a directory the agent does not watch.

Updating is the same command again. We refresh snapshots from GitHub on a regular schedule; the last-change date on the listing is the signal that the author changed the file, not repository stars.

FAQ

Do I install every skill on a task page? No. Pick one or two whose “when to use” matches the job. Overlapping taste files will compete on triggers. The choose a skill guide walks the common overlaps.

Is this an MCP server? No. A skill is instructions. MCP is a live tool connection. Some SKILL.md files mention MCP; enabling that is a separate decision in the agent.

Can I install from this website’s copy? Don’t, for Claude, Cursor, or Codex. Use GitHub via the command we print. Our page is a snapshot for reading; the author can change the file tomorrow. Pasting into Grok Bot is the exception: that product has no skills folder here, so the chat gets the text.

Can I use this with Grok Bot? Yes. Paste the SKILL.md or attach the file. Grok Bot instructions for one job, which are not frontend skills, live under Grok Bot playbooks.