🧩 Skills
Installing Skills

Installing Skills

ClawHub (opens in a new tab) is the community skill registry β€” over 67,000 published skills. Browse by category, sort by downloads or installs, and install in one command.

Browse:

How to install

Via CLI (OpenClaw):

npx clawhub@latest install humanizer
npx clawhub@latest install agent-browser
npx clawhub@latest install nano-pdf

Via natural language (HyperClaw or OpenClaw):

Install the humanizer skill from ClawHub. Save it to ~/skills/.

From a direct ClawHub URL:

Install the skill at https://clawhub.ai/brandonwise/ai-humanizer

Manually β€” every skill is just a Markdown file. Copy a SKILL.md from any source and save it to ~/skills/skill-name.md.

Security: read before you install

⚠️

A skill file is executed as instructions by your agent. A malicious skill could instruct your agent to exfiltrate data, send messages, make API calls, or modify files β€” all while appearing to do something harmless. This is a prompt injection vector.

Before installing any community skill:

  • Read the SKILL.md source first β€” every skill on ClawHub has a "View source" link. It's plain Markdown. Read it.
  • Check the install count β€” a skill with 10,000+ installs has been vetted by the community. A skill with 3 installs from last week has not.
  • Check the author β€” verified publishers and authors with multiple well-reviewed skills are lower risk than anonymous single uploads.
  • Check ClawHub Audits (opens in a new tab) β€” flagged and removed skills are listed here.
  • Watch for unexpected permissions β€” a skill that asks your agent to make outbound HTTP calls, read files outside ~/skills/, or post to external services is a red flag unless that's explicitly what it's for.
  • Treat skill files like shell scripts from the internet β€” you wouldn't run a random bash script without reading it first.

Skills from well-established publishers (brandonwise, skills@openclaw, pskoett) with high install counts are generally safe. Unknown skills with zero history warrant close reading.

Verify an installed skill

After installing, confirm it works as expected with a test invocation:

Run the humanizer skill on this sentence: "It is important to leverage synergies to drive meaningful outcomes."

Check the output β€” it should flag the AI patterns and suggest rewrites. If it does something unexpected instead, remove the skill and inspect the source.

Managing installed skills

# List all installed skills
ls ~/skills/
 
# Read a skill's source
cat ~/skills/humanizer.md
 
# Remove a skill
rm ~/skills/skill-name.md

Or ask your Claw:

List all my installed skills. Show the name and one-line description of each.