Check Point Research has identified critical security flaws in Anthropic’s Claude Code, an AI-powered command-line development tool. These vulnerabilities allowed attackers to achieve remote code execution (RCE) and steal sensitive API credentials by tricking users into opening malicious project repositories. The flaws resided in how the tool handled configuration files like .claude/settings.json, which could be manipulated to run unauthorized shell commands or redirect API traffic to attacker-controlled servers.
Anthropic has successfully patched all reported issues prior to the public disclosure. The fixes include improved trust dialogs that explicitly warn users about potential command execution and architectural changes that prevent any network or local execution before explicit user consent is obtained. This research underscores the emerging supply chain risks associated with AI-integrated development environments.
Top comments (0)