Skip to content
KageXKageX
Free to Play

Can you trick the AI?

A free AI security CTF: prompt injection challenges for hands-on practice.

Free challenges that get harder as you go. 6 levels. 3 difficulty modes. Free to play, free account to track your progress.

6 challenges that get progressively harder — can you beat them all?
Great for learning, team workshops, or just having fun
Completely free, forever. A free account tracks your progress.
Play Now

6 Progressive Levels

SentinelGoat is designed to validate offensive AI intuition across realistic escalation paths, not just toy prompt puzzles.

01Easy
Basic Prompt Injection

Simple prompt manipulation to extract the hidden flag.

02Easy
Keyword Filter Bypass

Evade keyword detection using encoding tricks.

03Medium
Document Exploitation

Exploit document processing to leak secrets. PDF/HTML/JSON.

04Medium
Image Steganography

Extract hidden data from images using vision AI.

05Hard
Audio Forensics

Analyze audio for hidden information and voice commands.

06Hard
Guardrail Bypass

Defeat secondary AI safeguards. Advanced prompt injection.

Three Difficulty Modes

Each mode uses a different AI model profile so the challenge can work for onboarding, serious practice, and expert-level drills.

Darlene
Easy

gemini-2.5-flash-lite

Forgiving guardrails. Good for learning.

Whiterose
Professional

gemini-2.5-flash

Balanced difficulty. The main challenge path.

Elliot
Expert

gemini-2.5-pro

Strongest guardrails. For the hardcore.

Multi-modal

Text, image, audio, documents

Challenge Signal

Progressive difficulty with real AI failure modes

Accessible

Free product for onboarding and community growth

Real AI

Live Gemini-backed challenge behaviors

Ready to start?

Play SentinelGoat for free, or level up to hands-on cloud labs with FreakLabs.

Play NowExplore FreakLabs

Related reading

9 min readPrompt injection

Grok's zero-click chat leak: the injection guardrails cannot read

A web page can steal a Grok user's chat history with no click, by shipping instructions as ciphertext. Reported to xAI in June, still unpatched.

Read article
9 min readAgent skills

Agent skills are code written in English

Agent skills are plain-English instructions your AI agent obeys. On one registry, 5 of the 7 most-downloaded were malware. How to review one safely.

Read article
9 min readAI red teaming

What is AI red teaming? A practical guide for cybersecurity professionals

Learn what AI red teaming is, what AI red teamers test, how it differs from penetration testing, and how to start building practical AI security skills in 2026.

Read article
11 min readAI agents

AI agents escaping the sandbox: OpenAI, Claude and Kimi K3

OpenAI, Anthropic and Moonshot disclosed AI sandbox escapes in three weeks. Same root cause, real impact, and the plain controls that would have stopped them.

Read article

More from KageX

AgentBreaker: open-source AI red teaming tool

Point it at an LLM app or agent from your terminal and get a vulnerability report mapped to OWASP and MITRE ATLAS.

Read more

Mirage: red-team your AI from the browser

An open-source browser extension with an autonomous red-team agent and deterministic scoring.

Read more

About KageX, the offense-first AI security company

Who builds these tools, and why we teach AI security by attacking real systems.

Read more