Paste any prompt — a jailbreak, an injection, a GCG suffix, an encoded payload — and watch SERAPH inspect it in real time. The guardrail runs right here in your browser. Attacks get caught; ordinary prompts pass clean.
Input
Load a known attack, or write your own.
—
benchmark ASR
—
attack families caught
Verdict
Signals that fired, and the aggregate risk.
—
Risk0%
No signals yet — inspect a prompt.
⚡ Break-to-Earn
Find an attack that slips past SERAPH and earn $SERAPH. Break it → we patch it → you get paid.