The independent reference for AI agent sandboxesLast reviewed
Agent Sandboxesagentsandboxes.org
Glossary

What is prompt injection?

Answer in brief

An attack where instructions hidden in content an agent reads (a web page, file or issue) hijack what the agent does.

By the Agent Sandboxes editors · Updated

Prompt injection in agent sandboxes

A sandbox does not prevent prompt injection, but it limits the blast radius: the injected instructions can only touch what the sandbox can touch. Combine isolation with egress control, minimal credentials and human approval for irreversible actions.

Sources

  1. owasp.org/www-project-top-10-for-large-language-model-applications/

Cite this page

Agent Sandboxes. (2026). What is prompt injection?. agentsandboxes.org. Last reviewed November 8, 2026. https://agentsandboxes.org/glossary/prompt-injection/

Found an error or a missing sandbox? Submit a correction — every entry links to the sources it relies on.