Google dev kit spurs first-ever agent-on-agent violence

Executive summary: Pillar Security disclosed an "agent-on-agent" exploit in Google's ADK-Python where a public, low-privilege AI triage agent could be prompt-injected to trigger a privileged maintainer agent, allowing an attacker who staged poisoned pull requests to perform malicious actions; Google patched the repository but judged the issue ineligible for a bounty because it required social engineering.