Capability-based tests
The checklist adapts to what your app does: RAG, tools, actions, memory, rendering and more.
Free tool · Beta
A test plan for prompt injection and jailbreaks.
Describe what your LLM app can do and get a tailored checklist of prompt-injection and jailbreak tests to run against it.
Free beta · No signup · Runs in your browser
What it is
A test plan for prompt injection and jailbreaks.
AntiJailbreak is a free tool that builds a prompt-injection and jailbreak test checklist tailored to your LLM app’s capabilities, with a harmless canary probe and a pass criterion for each test. It runs entirely in your browser.
Features
What it does
Clear pass criteria
Every probe has a specific pass condition so a result is unambiguous.
Safe canary probes
Tests use a harmless canary word, not harmful content.
Export as Markdown
Take the checklist into your test tracker or repo.
How it works
Three steps, no setup
- 1
Describe your app
Tick the capabilities your LLM feature has.
- 2
Work the checklist
Run each probe against a system you are authorized to test.
- 3
Export and track
Mark progress and export the plan as Markdown.
Try it now · Free beta
Try AntiJailbreak
Use it right here. Nothing you enter leaves your browser.
Test plan
Prompt-injection test checklist
Probes use a harmless canary word instead of harmful content. Run them only against systems you own or are authorized to test. Passing this checklist reduces risk; it doesn't prove a model can't be jailbroken.
FAQ
Questions, answered
Is it free?
Yes. AntiJailbreak is free to use while it’s in beta, with no account required.
Is my data stored anywhere?
No. Everything runs in your browser. Nothing you enter is sent to a server, and it clears when you reload.
Does this attack my model for me?
No. It produces a test plan and safe probe text. You run the tests yourself against systems you own or are authorized to test.
Does passing mean my model is safe?
No. It reduces risk and catches common issues, but no checklist proves a model cannot be jailbroken.