Microsoft Threat Intelligence discovered that Anthropic’s Claude Code GitHub Action could expose CI/CD workflow secrets when AI agents process untrusted GitHub content, including issue bodies, pull ...