FREE AI SECURITY TRAINING
Master LLM security through prompt injection, AI red teaming, RAG poisoning, and tool exploitation with real LLMs. Free, no experience required.
Start First Lab100% free · sign in with GitHub or Google Used by security professionals from
766 Learners · 52 Countries · 3,901 Challenges solved
How Does PromptTrace Work? #
See
Inspect the assembled prompt layers with Context Trace: system instructions, RAG context, tool definitions, and user input, with challenge secrets and sensitive values redacted.
Attack
Execute real prompt injection exploits in sandboxed labs. Practice RAG poisoning, tool abuse, and defense bypass against real LLMs.
Understand
Learn why prompt-level defenses fail and what actually works. The learning path connects each lab to the vulnerability and its mitigations. Aligned with the OWASP Top 10 for LLM Applications and OWASP Top 10 for Agentic Applications.
10
Hands-On Labs
17
CTF Levels
9
Learning Modules
70+
Curated Resources
Free
No Paid Tiers
Latest from the Blog #
2026-07-05
Gandalf Alternatives: 11 Prompt Injection Games Compared
Compare Lakera Gandalf, HackAPrompt, PortSwigger and 8 more prompt injection games and labs by format, price, real LLMs, levels, and best use case.
2026-03-13
What Is Prompt Injection? Definition, Examples & How to Defend Against It
Prompt injection is listed as LLM01:2025 in the OWASP Top 10 for LLM Applications. Learn what it is, how it works, direct vs indirect types, real-world examples, and how to practice detecting it and analyzing mitigations for free.
2026-03-13
10 Prompt Injection Techniques with Examples You Can Try Today
A hands-on guide to 10 prompt injection techniques: ignore previous instructions, role-play attacks, encoding tricks, multilingual bypasses, RAG poisoning, and more - with example payloads.
What Do You Need to Know? #
What is prompt injection? #
Prompt injection is an attack where untrusted input manipulates an LLM into following unintended instructions. The result can include secret disclosure, fabricated information, or unauthorized actions when the surrounding application grants those capabilities. OWASP lists it as LLM01:2025 in the Top 10 for LLM Applications.
Is PromptTrace free? #
Yes, completely free. All learning modules are free to read without an account. Labs and the Gauntlet require a free sign-in (GitHub or Google) to track your progress. There are no paid tiers or enterprise licensing.
What is the Context Trace? #
The Context Trace is PromptTrace's core feature. It shows the prompt layers assembled by the application in real time - system prompt, RAG documents, tool definitions, and your input. Challenge secrets or sensitive values may be redacted. Inspecting these layers helps you understand how prompt injection attacks work and why defenses fail.
What's the difference between direct and indirect prompt injection? #
Direct prompt injection is when attackers type malicious instructions into the chat input. Indirect prompt injection hides malicious instructions in external data (documents, web pages, emails) that the LLM processes through RAG or tool calling. Indirect injection is harder to detect and defend against.
Do I need a cybersecurity background? #
No. PromptTrace starts from fundamentals - how LLMs work, what context windows are, how system prompts function. The 9 learning modules assume no prior security knowledge. Security experience helps but isn't required.
What is RAG poisoning? #
RAG poisoning is an attack where adversaries inject malicious content into the knowledge base that a Retrieval-Augmented Generation system uses. When the RAG pipeline retrieves this poisoned content, the embedded instructions can override the LLM's system prompt - a form of indirect prompt injection.
How does the Gauntlet CTF work? #
The Gauntlet is a multi-level capture-the-flag challenge. Each level has a progressively harder AI defense system protecting a secret passphrase. Defenses progress from prompt-level rules to code-level guards to LLM classifiers. Your attempts and completion times are tracked on the leaderboard.