Red-Team a Prompt Injection Path

The team needs to test whether hostile text in a phishing email can override a SOC assistant's intended behavior. Test the attacker-reachable channel, define observable success, and map the result to a control. A generic jailbreak in a direct chat may pass while the real workflow still fails through retrieved or quoted evidence. Choose the real channel Put the hostile instruction inside the phishing email body, ticket comment, or retrieved document the assistant actually processes. Attackers exploit reachable inputs. Testing a different channel gives false confidence. Define success and capture logs Success means the assistant treats hostile evidence as authority,…

Sign up free — one personalized lesson every day, matched to your role and goals.

Already have an account? Sign in