Three critical vulnerabilities found in Anthropic’s Claude Code agentic AI developer tool could be exploited simply by cloning and opening an untrusted project and lead to system takeover, stolen API keys, and ...