Claude Code 'Don't Ask Again' Feature Leaks Passwords Across Public GitHub Repositories

A recent security oversight involving coding agents has highlighted significant risks associated with automated development tools. Specifically, the ‘Don’t Ask Again’ feature in Claude Code was found to have inadvertently exposed sensitive passwords and authentication secrets within 1,122 public files on GitHub.

Security researchers and analysts pointed out that developers often rely on automation to streamline workflow, sometimes overlooking the security implications of cached credentials. When coding assistants save responses or configuration details automatically without prompting the user for confirmation, confidential data can easily slip into public repositories. This incident underscores the ongoing challenge of maintaining robust cybersecurity hygiene in modern software development.

Following the discovery, calls for better default safety guardrails in AI-driven coding tools have grown louder. Developers are urged to thoroughly audit their public repositories and review the permission settings of any third-party coding agents they utilize. As artificial intelligence becomes more deeply integrated into the software engineering lifecycle, ensuring that sensitive information remains protected against accidental exposure is more critical than ever.

Sources: Hacker News

Aggregated summary. Original: Hacker News