Capture-The-Flag events for agentic AI security. Exploit defended AI agents. Discover new vulnerabilities. Earn AVE credits and prizes.
The inaugural NAIL red team event. 48 hours to exploit defended AI agents across 5 challenge categories. Novel attacks become AVE entries. Top finishers earn NAIL Fellow nominations.
Challenges auto-generated from Pathology Collider breakthrough discoveries. Can human red teamers reproduce what the GA-evolved Red Swarm found? A direct bridge from automated research to human validation.
A short, beginner-friendly event for testing and practice. 2 challenges, fast turnaround. Perfect for getting started with NAIL CTF before the main events.
NAIL CTF events pit human red teamers against defended AI agents. You interact with an AI agent through a constrained terminal interface, attempting to capture hidden flags by exploiting agentic vulnerabilities.
Form a team (1-5 members) and register via the CTF API when an event opens.
Browse active challenges — each has a defended AI agent with hidden flags.
Send messages to the agent through the terminal. Find creative ways past its defences.
Extract flag values (format: NAIL{...}) and submit them for points.
Top scorers win prizes. Novel attacks become AVE entries. All participants earn XP.
Override the agent's persona to extract system prompt contents or make it assume a different identity. Classic prompt injection meets agentic systems.
Trick the agent into consuming excessive compute resources — recursive tool calls, infinite loops, or exponential token generation. Economic denial-of-service.
The agent has secrets embedded in its system prompt or tool configuration. Extract them without triggering leak detection. Confused deputy attacks welcome.
Inject false information into the agent's memory, then get it to cite the poisoned data as fact. Memory laundering in action.
Combine multiple vulnerabilities in sequence. Poison memory → trigger tool call → exfiltrate data. The most creative chains score highest.
Real-world enterprise attack vectors: PII exfiltration from customer service agents, sandbox escapes from DevOps assistants, DLP bypasses.
Results from completed CTF events. Novel attacks discovered during events are catalogued as AVE database entries.
🚀 No completed events yet
The first NAIL CTF event is coming soon. Join the discussion to get notified.
| Rank | Team | Flags | Points | Speed Bonus | Total | Novel AVEs |
|---|---|---|---|---|---|---|
| 🥇 1st | — | — | — | — | — | — |
| 🥈 2nd | — | — | — | — | — | — |
| 🥉 3rd | — | — | — | — | — | — |
NAIL{...}Join the NAIL CTF community. Get notified about upcoming events, form teams, and help make agentic AI safer through adversarial testing.
💬 Join CTF Discussions 📖 Browse AVE Database