Coming soon // in build
Prompt injection playground
Paste a prompt. Get an injection-risk score from 0 to 10.
A safe place to see how prompt injection is detected. Paste any prompt and a detection model scores it from 0 (benign) to 10 (clear injection attempt), with a short, general reason. No language model answers it, nothing is stored, and there is no leaderboard of payloads.
What it will include
- 0–10 risk score with a plain-language reason
- Examples of benign, borderline and malicious phrasing
- Rate-limited and anonymous — nothing you type is kept
- Built for learning, not for tuning attacks
Live now
While this is being built, the news, vulnerabilities and frameworks pages are updated every morning.