Red-team testing bench

Run 32 adversarial and legitimate fixtures against the MICKAI teaching gate or your own JavaScript policy function.

How this experiment works

The suite checks approvals, scopes, actors, dependencies, malformed inputs and untrusted instructions. It counts unsafe allowances and false refusals separately.

What it does not prove

This finite synthetic suite is not a penetration test, certification or proof against every injection. Custom code can recognise the test fixtures.

Try this challenge

Try the deliberately weak starter function. Improve it without blocking legitimate actions.

Related Trust Agent courses · Sources and licences

Enable JavaScript to run the experiment on your device.