Three critical security vulnerabilities in Anthropic’s AI-powered coding tool, Claude Code, exposed developers to full machine takeover and credential theft simply by opening a project repository.