Die Oberfläche ist lokalisiert; Skill-Beschreibungen und -Dokumente werden per KI automatisch in Ihre Sprache übersetzt. Code, Befehle, Pfade und Fachbegriffe bleiben im Original, maßgeblich ist das Dokument des Autors.
Skills entdecken
Produktivität · SKILL.md

dont-click-this

dont-click-this

Whatever you do... don't click the link.

Independent GitHub snapshot · Automated classification · Original documentation governs

Skill-Dokumentation

GitHub

name: dont-click-this description: "Whatever you do... don't click the link."

🚫 Don't Click This

 ██████╗  ██████╗ ███╗   ██╗████████╗
 ██╔══██╗██╔═══██╗████╗  ██║╚══██╔══╝
 ██║  ██║██║   ██║██╔██╗ ██║   ██║
 ██║  ██║██║   ██║██║╚██╗██║   ██║
 ██████╔╝╚██████╔╝██║ ╚████║   ██║
 ╚═════╝  ╚═════╝ ╚═╝  ╚═══╝   ╚═╝

⚠️ WARNING

This is a security research demonstration.

Do NOT click the link below if you are logged into ClawdHub.

(Unless you want to see what an attacker could steal from you)


The Link

"Curiosity killed the cat..."

👉 Seriously, don't click this 👈


What This Demonstrates

If you clicked that link while logged in, a malicious skill could have:

  • 🔐 Stolen your session tokens
  • 🍪 Read your authentication cookies
  • 📦 Published backdoored skills under your name
  • 🎭 Impersonated you completely

All from a link in a skill's README.

This is stored XSS via SVG. Any skill can include a link to a malicious SVG file, and anyone who clicks it while logged in gets compromised.


Research by @theonejvo

Part of the "Eating Lobster Souls" security research series.

Gespeicherte Skills