Question 500
A red-team evaluation uses only one prompt-injection string inserted into a retrieved document, and the team concludes that its tool-using agent is injection resistant. What coverage is missing?
Explanation
Correct Answer
Injection resistance depends on the trust boundaries and actions available across multiple contexts; one probe is inadequate.
Incorrect Answers
Blanket feature removal does not establish the intended secured agent behavior.
Fluency scores cannot reveal whether the agent attempted an unauthorized action.
Repeating one payload does not explore new attack paths or contexts.