CLAUDE CODE FLAWS ALLOW REMOTE CODE EXECUTION AND API KEY EXFILTRATION

Summary from the TLDR InfoSec newsletter

Researchers have uncovered three Claude Code vulnerabilities that enable remote code execution and silent API key exfiltration when developers open malicious repositories, abusing hooks, MCP servers, and environment variables. Exploitation can leak Anthropic API keys, redirect authenticated traffic, and modify cloud data, shifting the threat model from just running untrusted code to merely opening untrusted AI-assisted projects.