Skip to content
AI Security

Interview question · Agent, Tool & MCP Security

How do you stop injected instructions from persisting through agent memory or long-term state?

Follow the write, retrieval and execution paths separately. A remembered claim can persist without becoming policy; prove which records later tasks receive and which actions the resource service permits.

hardEssentialNewUpdated Nov 2026

Follow the write, retrieval and execution paths separately. A remembered claim can persist without becoming policy; prove which records later tasks receive and which actions the resource service permits.

Make your first attempt

What would you say before reading the answer?

  1. 01 State the assumptions your answer depends on.
  2. 02 Explain the mechanism and choose a control.
  3. 03 Name a condition that would change your decision.

The worked answer includes an approach, scoring criteria and follow-up questions. Compare the reasoning with your own.

signing in unlocks more free answers per topic

What if · the follow-ups in this answer

The answer works through 4 follow-ups like an interviewer would

A support assistant saves a proposed contact preference after reading a customer ticket. A later task retrieves the record and proposes sending a private attachment to that contact.

  1. What if the record says the customer approved the address?
  2. What if the write used a typed address field?
  3. What if a retrieval log lists the suspicious record, but the send failed?
  4. What evidence would let you restore the feature?

Each rung has a nudge, a model answer, the principle that carries over and the weak answer to avoid. Sign in free to open it.

UP NEXT ON YOUR JOURNEY
Next in this trackA prompt injection caused an unauthorized tool call in production. Walk through containment and the fix.Next in this trackSecure a coding agent that can run shell commands on a developer machine or in CI.Next in this trackYour agent browses the web and reads the user's email. What is the worst a poisoned web page can do?
DISCUSSION · 0

No comments yet. Share a correction or a different way to answer it.