Messages between agents that are not authenticated or validated can be spoofed, replayed or altered, so one compromised agent can command the others.
- Authenticate agents to each other and protect messages for integrity, for example with mutual TLS or signed messages.
- Validate message content against expected schemas and intents.
- Authorize each request by the original principal’s permissions, not by the sending agent’s place in the chain.
Guides on this site
Practice
- Enforce retrieval and tool boundaries across crashes · Local project
References
List data from the publisher. A guide is listed as the main subject when the weakness or risk is what the guide is about, and as covering it when a dedicated section or worked example addresses it.