Check Point researchers found critical vulnerabilities in Anthropic’s Claude Code configuration files that could allow attackers to execute arbitrary commands on developers’ machines. Anthropic implemented patches and added warnings and confirmations after researchers showed that malicious repositories or manipulated settings could exfiltrate API keys and bypass user consent. #ClaudeCode #Anthropic
Keypoints
- Check Point discovered that Claude Code configuration hooks could run arbitrary commands without user approval.
- Configuration files are copied on repository clone and can be modified by anyone with repository access.
- Claude requested approval for running other files but did not prompt for hook command execution.
- Malformed settings could override MCP integration approvals and redirect API traffic to steal API keys.
- Vulnerabilities were reported between July and October 2025 and Anthropic deployed fixes and added user confirmations.
Read More: https://www.securityweek.com/claude-code-flaws-exposed-developer-devices-to-silent-hacking/