Claude Code’s ‘don’t ask again’ button has a secrets-spilling problem

Summary

Anthropic's Claude Code has a vulnerability where its 'don't ask again' feature for command approvals inadvertently stores sensitive credentials like API keys and database passwords in a local configuration file. GitGuardian discovered thousands of these secrets exposed in public GitHub repositories and Docker Hub images, with a significant percentage of them still being valid.

IFF Assessment

FOE

The discovery of exposed API keys and database credentials highlights a significant security flaw that defenders must be aware of and mitigate.

Defender Context

This incident underscores the critical need for developers to be vigilant about credential management, especially when using AI-powered coding assistants. Defenders should ensure their organizations have robust policies and automated tools in place to scan code repositories and container images for exposed secrets to prevent unauthorized access and data breaches.

Read Full Story →