Prevent Email Agent Abuse

Scenario

You are building an agent that can draft and send emails on a user's behalf. The feature is useful for follow-ups, scheduling, and customer communication, but it could also be abused to send spam, impersonate people, or generate phishing messages.

Question

How would you prevent the agent from being weaponized for spam or phishing while keeping it useful for legitimate email tasks?

Problem

Scenario

Question

How would you prevent the agent from being weaponized for spam or phishing while keeping it useful for legitimate email tasks?

What this tests

Agent guardrails for high-risk actions
Prompt injection awareness from untrusted context
Hallucination and deception containment
Eval design for safety-critical LLM features

Problem

Scenario

Question

How would you prevent the agent from being weaponized for spam or phishing while keeping it useful for legitimate email tasks?

What this tests

Agent guardrails for high-risk actions
Prompt injection awareness from untrusted context
Hallucination and deception containment
Eval design for safety-critical LLM features

Problem

Scenario

Question

How would you prevent the agent from being weaponized for spam or phishing while keeping it useful for legitimate email tasks?

What this tests

Agent guardrails for high-risk actions
Prompt injection awareness from untrusted context
Hallucination and deception containment
Eval design for safety-critical LLM features

Interview Guides

Problem

Scenario

Question

What this tests

Problem

Scenario

Question

What this tests

Prevent Email Agent Abuse

Problem

Scenario

Question

What this tests

Problem

Scenario

Question

What this tests