Flaws in Google APK for Python Unlock Agent-to-Agent Attack

Researchers at Pillar Security discovered a critical vulnerability in Google's Agent Development Kit (ADK) for Python, allowing a low-privileged AI agent to trigger actions by a more privileged agent via prompt injection within GitHub pull requests. This represents a new attack vector targeting the emerging trend of using AI agents in software supply chains, highlighting a significant shift in attack surfaces and demanding a reevaluation of security models for AI agent workflows. Google has addressed the flaws, but the discovery underscores the need for organizations to carefully govern agentic workflows and restrict agent access to prevent potential supply chain compromises.