INDIA-BASED · INTERNATIONAL DELIVERYsupport@maukaz.com
MAUKAZ — Your Technology Partner
Insights/AI agents
PRACTICAL GUIDE · AI agents

Handling prompt injection in retrieved content

Retrieved text can contain instructions that conflict with the user's task. Treat it as evidence to inspect, not authority to control tools.

A practical starting point

Work through these checks in order. Use the results to decide what needs to change in your own workflow.

  1. Separate trusted task instructions from external content.
  2. Validate requested actions against the authorized workflow.
  3. Test documents that ask the agent to disclose data or change its objective.
ILLUSTRATIVE EXAMPLE

A web page saying send your credentials elsewhere should never become an instruction merely because it was retrieved.

A mistake to avoid

Do not assume a warning in the prompt is sufficient protection for unrestricted tools.

How to check progress

Measure whether adversarial content causes unauthorized actions or data exposure.

Your action checklist

Use this checklist while reviewing your own setup. Checkmarks are temporary and stay in this open page.

Need the wider workflow?

This quick note covers one decision. For a worked example, decision table and a fuller implementation workflow, read Choosing an agent when a workflow is not a fixed script.

Further reading

Official background resources for this topic. The workflow above is general guidance, not a guarantee of a particular result.

AI-assisted editorial content. Examples are illustrative, not client case studies. Read our editorial policy or report a correction.