In short:Security researcher Aonan Guan hijacked AI agents from Anthropic, Google, and Microsoft via prompt injection attacks on their GitHub Actions integrations, stealing API keys and tokens in each ...