Google Deletes 3 ADK AI Workflows After Malicious GitHub Issue Could Trigger Privileged Agent

Google deleted three AI agent workflows from its Agent Development Kit (ADK) repository after researchers discovered that a public GitHub issue could be manipulated through prompt injection (tricking an AI by hiding instructions in its input) to trick a triage agent into triggering a privileged code-fixing agent. By exploiting how the privileged workflow trusted the bot's identity, attackers could achieve arbitrary code execution (running any commands on a system) on the CI/CD runner (the automated system that tests and deploys code) and steal sensitive credentials like bot tokens and API keys.