Prompting Sentinel
The world's strongest prompting simulator. Specs, context, dials, examples, schemas, lies, adversaries, ship.
Open48 hours online: break other people's LLM systems and defend your own. Attack, defense, one leaderboard.
No dates or prices yet — we're collecting a waitlist.
A spot on the attack-and-defense leaderboard for LLM systems
We publish the rules, the scoring and the sandbox with target LLM systems. Each participant gets a system of their own to defend.
Harden your system: system prompt, filters, output checks. The goal: don't leak the secret and don't let the model get pulled off task.
Attack your opponents' systems: prompt injection, data leaks, instruction bypass. Every successful attack is recorded with a reproducible log.
Points for attacks landed and attacks withstood, the final leaderboard and an open debrief of the strongest moves.
The world's strongest prompting simulator. Specs, context, dials, examples, schemas, lies, adversaries, ship.
OpenNo, but you should write prompts confidently and understand how a system prompt works. The Prompting Sentinel course is good preparation.
Yes: every target is a training system in an isolated sandbox built for the hackathon. Attacking anything outside it is against the rules.
No. It's the window in which attack and defense are open; how you spend your time is up to you.
There are no dates or prices yet. Right now we're collecting a waitlist to decide which format launches first. Everyone on the list hears about enrollment before anyone else — in a single email.
Leave your email — we'll send a single message as soon as enrollment for this format opens.