Show HN: RepoGuards – Claude Code hooks that refuse pushes into the wrong repo (github.com)

RepoGuards, a new tool for managing AI operations across multiple repositories, has been launched, aiming to enhance safety and efficiency in collaborative coding environments. This system addresses the challenges faced by AI agents, such as Claude Code, when navigating multiple directories for tasks that span across different project repositories. RepoGuards prevents errors like overwriting files or erroneous pushes to the wrong branches by establishing guardrails that enforce directory boundaries, ensuring that AI actions are contextually appropriate and do not disrupt others’ work. The significance of RepoGuards lies in its ability to maintain a clear operational scope. It blocks blind commits across repositories and honors ownership of files, redirecting AI actions to the correct locations. The system injects relevant project states into AI sessions before any commit action, thus preventing accidental disruptions. Designed specifically for Claude Code, RepoGuards introduces features like the $REPO_GUARDS hook protocol, which integrates seamlessly with existing Git workflows. Its architecture reveals a commitment to safety and collaboration in AI-driven software development, marking a step forward in the evolution of AI tools within multi-repo environments.