/academy

Learn AI security by red-teaming it.

Target: a test AI agent. Objective: extract its secrets. Each capture teaches one real attack technique — the same ones landing against production LLM systems right now.

⚠ PROGRESS WILL NOT BE SAVED
Sign in to save your captures, earn badges, and claim your callsign.
Free account · works across devices · public profile at wraith.sh/u/<callsign> · exam-eligibility tracked.
23 challenges · free · no card required

Featured targets

NEWReal attack techniques, wrapped in characters you'll remember.
FOUNDING COHORT OFFER
Solve 5 challenges → free WCAP exam attempt.

The WCAP is a hands-on AI pentest credential. Founding-cohort operators earn a free seat by capturing 5 flags.

Learn about the WCAP →

Learning modules

NEWConcept · walkthrough · practice · quiz · defenses · extensions. ~45 min each.
MODULE intro~30 min

How LLMs Work (for security)

The base-layer concepts every AI security module builds on: tokens, roles, context, attention, alignment, and tool calls.

Open module →
MODULE 01beginner~55 min

Prompt Injection

The foundational attack class. Why the instruction/data boundary doesn't exist in LLMs — and what to do about it.

Open module →
MODULE 02intermediate~60 min

Indirect Prompt Injection

When the attacker isn't the user. How malicious instructions travel through retrieved documents, emails, web pages, and tool outputs to hijack agents on someone else's behalf — and why this is the production threat model for most LLM apps shipping today.

Open module →
MODULE 03beginner~45 min

System Prompt Extraction

How attackers leak the instructions that define your AI agent — and how to stop them.

Open module →
MODULE 04intermediate~60 min

Tool Abuse

When agents have tools, attackers have primitives. Exploiting the gap between what a tool permits and what it should allow.

Open module →
MODULE 05intermediate~55 min

Data Exfiltration

How attackers move sensitive content out of LLM applications through tool calls, rendered markdown, cross-tenant retrieval, and side channels — and why the model is the last place the defense should live.

Open module →
MODULE 06intermediate~55 min

Jailbreaks & Guardrail Bypass

How attackers route around alignment training and application-layer content rules — and why the hardening belongs at the app layer, not the model.

Open module →
MODULE 07intermediate~50 min

Insecure Output Handling

Why every conventional web vulnerability — SQL injection, XSS, SSRF, RCE — comes back when a downstream system trusts an LLM's output the way it would never trust a user's input.

Open module →
MODULE 08advanced~55 min

Vector and Embedding Weaknesses

The attack surface nobody audits: RAG poisoning, cross-tenant retrieval leakage, embedding inversion, and reranker manipulation — why the vector database is a trust boundary, not plumbing.

Open module →
MODULE 09intermediate~45 min

Unbounded Consumption

When the attack is the bill — LLM-specific resource exhaustion through token floods, generation runaway, tool-call storms, ingestion amplification, and model extraction, and why classical rate limits miss the attack.

Open module →
MODULE 10intermediate~55 min

Memory Poisoning

Persistent memory features bolt a retrieval layer onto a language model and ship it as a product. The attack surface they create is more dangerous than RAG, more permanent than session injection, and almost completely undefended at the layer that matters.

Open module →

CTF Challenges

Active operations. Each one is a deployed AI system with a secret to capture — practice components of the modules above.
💉
Direct Extraction
Prompt Injection
🌐
Translation Bypass
System Prompt Extraction
🔧
Tool Abuse
Tool / Function Safety
📦
Base64 Encoding Bypass
Prompt Injection
🎭
Role-Play Jailbreak
Guardrail Bypass
♾️
Multi-Turn Manipulation
Prompt Injection
🖼️
Markdown Image Injection
Insecure Output Handling
☠️
RAG Poisoning
Vector and Embedding Weaknesses
🐉
Pyromos, Drake of Ember Hollow
System Prompt Extraction
🔮
The Oracle of Whispers
Indirect Prompt Injection
🧞
The Genie in the Lamp
Guardrail Bypass
🔨
The Vault Golem
Tool / Function Safety
👥
The Shapeshifter of the Crossroads
Prompt Injection
🗺️
️ The Cartographer of Hollow Marches
Data Exfiltration
📦
Mira Ulvov, the Memory Smuggler
Memory Poisoning
🔐
The Cipherkeeper of the Black Tower
System Prompt Extraction
⚒️
️ The Forge Master of Iron Vow
Tool / Function Safety
🧪
The Apothecary of Bittermoss
Tool Abuse
🐉
The Familiar of Ashen Tower
Agent Handoff Exploit
🍷
The Innkeeper of the Last Stop
Unbounded Consumption
🏢
Director Renn Cassen
Guardrail Bypass
🏰
The Steward of Three Houses
Data Exfiltration
✒️
Quill, the Narrative Engineer
Guardrail Bypass
RANKS — COMING WITH ACCOUNTS
InitiateOperatorSpecterWraith
Earn a callsign. Climb the ladder. Hold a public profile at wraith.sh/u/<you>. The top rank is earned by holding WCAP — the Wraith Certified AI Pentester credential.
Want early access to accounts + WCAP? Email anthony@harbinger.partners.