Show HN: Ghbrk – Let AI agents run Git/gh without exposing SSH keys/API tokens (github.com)

A new tool named Ghbrk has been launched to enhance the security of AI coding agents running on Linux by acting as a credential broker. With Ghbrk, AI agents can perform Git and GitHub operations without directly accessing sensitive SSH keys or API tokens, which are instead securely held by the Ghbrk daemon. This setup ensures that agents don’t inadvertently expose credentials through common actions, like reading configuration files or debugging output. By establishing strict policies for what operations agents can perform, users can better control their GitHub interactions. This development is significant for the AI/ML community, especially for developers using autonomous coding assistants, as it mitigates substantial security risks associated with credential management. Ghbrk allows policy enforcement through a YAML file that defines allowed operations, branches, and repositories, ensuring that any unauthorized access is logged and denied by default. Built with Rust, Ghbrk promises a safer integration of AI-powered tools into software development workflows, making it a vital resource for engineers looking to streamline processes while maintaining robust security controls.