Research published on August 3, 2026 describes a prompt injection attack that used a low-privilege agent to trigger a maintainer-only agent in Google’s open-source Agent Development Kit for Python.
verifying reliability
No specialized terms available for this topic.