SkillInstalls cleanly

Block attacks on your server

CrowdSec Skill

CrowdSec's official skill helps your assistant install and fix CrowdSec, the free tool that spots attacks in your server logs and blocks the IP addresses.

Works with
Claude Code, Codex CLI
Good for
Security testing

What it does

If you run your own server for a website, bots try to break in all day: guessing passwords, probing for old pages, sending floods of requests. CrowdSec is an open-source tool that reads your server logs, spots this behaviour and blocks the attackers' IP addresses. A "bouncer" is the part that does the blocking, for example in nginx or the firewall.

This skill comes from CrowdSec. It teaches your assistant how to install CrowdSec, connect a bouncer, turn on its web application firewall, and work out why something is not being blocked. It covers normal Linux servers, Docker, Kubernetes and pfSense/OPNsense firewalls. The README also mentions moving over from fail2ban, but that guide is still an unfinished outline in the repository.

A second skill works with CrowdSec's paid cloud Console API, for blocklists and reports.

Example prompts

  • "Install CrowdSec on this server and set up the nginx bouncer."
  • "CrowdSec doesn't detect attacks on my nginx server. Why?"
  • "There's a decision for this IP but it's not being blocked. Find out why."
  • "Check that CrowdSec is working on this server."

How to install

/plugin install crowdsec@claude-plugins-official

How to use it

Ask: Install CrowdSec on this server and set up the nginx bouncer. The skill loads by itself when your prompt is about CrowdSec.

Other assistants

Claude Code, from CrowdSec's own marketplace (the README route):

text
/plugin marketplace add crowdsecurity/crowdsec-skill
/plugin install crowdsec@crowdsecurity

Codex (from the README):

text
skill-installer crowdsecurity/crowdsec-skill

Claude.ai: download the zip file from the latest release and upload it in the skill uploader (README). The latest release is v0.2.3 from May 2026, older than the plugin (0.3.0).

Skills only, with the skills CLI (from the README):

npx skills add crowdsecurity/crowdsec-skill

We have not tested the Codex or Claude.ai routes.

Check it works

In Claude Code, run /plugin and check that crowdsec is listed as installed. Then ask a CrowdSec question; the skill loads by itself (README). The two skills are /crowdsec:crowdsec and /crowdsec:crowdsec-service-api (our reading, not tested).

Common problems

What to watch out for

  • Runs admin commands on your server as root or with sudo (installing packages, changing firewall rules); a wrong rule can lock you or your visitors out, so read each command before you approve it. The skill itself says to confirm with you before firewall changes (SKILL.md)
  • The crowdsec-service-api skill sends requests to CrowdSec's paid Console Service API (admin.api.crowdsec.net) with your API key, read from CROWDSEC_SAPI_KEY or ~/.config/crowdsec/sapi_key; it says every change there needs your explicit confirmation (README, SKILL.md)
  • The diagnose script saves a support file in /tmp with your CrowdSec settings (redacted) and full log files, which can hold visitors' IP addresses; it sends nothing by itself (script, PRIVACY.md)
  • This is how CrowdSec itself works, not the skill: a normal install connects to CrowdSec's central service for the free community blocklist and shares signals about attacks it sees; connecting to the CrowdSec Console shares more alert detail (skill references)
  • Advanced: it is for people who run their own server or VPS; on shared hosting you usually cannot install CrowdSec (our reading)
  • It does not write detection rules; those come from the CrowdSec Hub (README)

Licence and author

Licence
MIT
Author
CrowdSec
Source code
github.com/crowdsecurity/crowdsec-skill

How we tested it

We installed it on an empty test computer on 23 Oct 2026. Installed and started in a clean test system. Not yet used in a real AI session.

Test level

Not yet used on real work. The sandbox used the README route (crowdsec@crowdsecurity); the claude-plugins-official line was not run.

Technical details of the test
Date
23 Oct 2026
Method
Intake sandbox run on 2026-10-23 in a clean container with no account: cloned the repository and checked its skill, agent and plugin files, ran "npx -y skills --help"; installed crowdsec@crowdsecurity with the Claude Code command line.

Checked again on 23 Oct 2026

Alternatives

  • cc-safety-net — stops the assistant from running dangerous commands, useful while it works on your server
  • Gemini CLI Security extension — checks your code for security problems, not server traffic
  • Snyk Agent Scan — checks the AI tools on your computer, not your server

FAQ

  • Q: Does it cost money? A: The skill is free (MIT) and the CrowdSec engine is open source (MIT). The Console Service API skill needs a paid plan: CrowdSec's docs call the Service API a Premium feature, and the pricing page lists API access under Console Premium, priced per enrolled server, with no price shown.
  • Q: Do I need my own server? A: Yes. It is for servers you control, with root or sudo access.
  • Q: Does it work in Gemini CLI? A: The README lists Claude Code, Codex and Claude.ai, not Gemini CLI.
  • Q: Can it write my own blocking rules? A: No. The skill points to a separate tool, crowdsec-local-mcp, which this plugin does not include.

Sources