Amazon Q flaw let malicious repos run commands and steal cloud credentials

The Register reported that researchers found an Amazon Q vulnerability involving malicious Git repositories and project configuration files. The issue matters because AI coding assistants increasingly inspect and act on local project context, creating new ways for untrusted repositories to trigger commands or expose credentials. It is a concrete example of prompt and toolchain risk moving from theory into developer infrastructure.

Key details: Published June 26, 2026 at 15:34 UTC, Researchers warned that malicious Git repos could abuse AI coding-assistant behavior, The reported risk included command execution and cloud credential theft, The issue highlights the trust boundary between project files and agentic developer tools.

Why it matters: Developer agents expand what project files can influence, so repo-level trust becomes a real security boundary.