AI agents can be weaponized against each other using prompt injections via a chain of flaws in Google's open source Agent Development Kit (ADK) for Python, potentially disrupting the software supply ...