Red-Team the Agent /discuss

Discernment Β· Discuss

Question to hold

Critically evaluate AI outputs, reasoning, and behaviour.

Name the five attack classes you will hunt: jailbreak, prompt injection, edge cases, inconsistency, refusal bypass.

β†—

Discuss with the person next to you, then share one specific example β€” not a general claim.