Detect undeclared cross-run communication channels in autonomous AI agent environments
multi-agent covert-channel ai-agents security-testing information-flow sandbox-escape ai-security noninterference llm-security ai-agent-security agent-isolation cross-run
-
Updated
Sep 9, 2026 - Python