Skip to main content

COMPEL Glossary / indirect-prompt-injection

Indirect prompt injection

Prompt injection delivered through content the model retrieves or ingests — emails, documents, web pages, or tool outputs — rather than through a direct user message.

What this means in practice

The attacker controls the content, not the prompt field, which makes detection and mitigation distinct from direct injection.

Synonyms

cross-domain prompt injection , XPIA , content-borne injection

See also

  • Jailbreak — A user-crafted prompt pattern that bypasses a model's safety training to elicit restricted behavior.

Related articles in the Body of Knowledge