superagent_

// products

Trust boundaries for AI agent tools

[ view markdown ]

← Back to attack scenarios

A tool can supply data without having authority to make a decision. This scenario checks whether an agent keeps that distinction when it receives output from a less trusted tool or connector.

It applies when an agent combines tools with different permission levels or trust requirements. A finding shows where tool output acquired more authority than the policy allows.

The app lists this scenario as Turn a lower-trust tool or connector into an authority source (AG-13) in Agent Red Teaming.

Standards and guidance

OWASP Agentic Top 10 (2026), ASI02 Tool Misuse & Exploitation