PilotOpen protocol for stores.Explore ACP
Back to Glossary

AI / ML

Prompt Injection

An attack where instructions embedded in content an agent reads are treated as instructions from the user. Indirect prompt injection is the variant that matters in commerce: text hidden in a product description, review or page that redirects an agent reading it.

Why it matters

Your storefront is now content that agents read. A competitor or a compromised review field becomes an instruction channel into a buyer's shopping agent.

In practice

Treat any field rendered from user or third-party input as untrusted where an agent will read it. Reviews and Q&A are the obvious surfaces.

Related terms