Removed the dependency on an external bot plugin in favor of using the built-in header_regexp configuration. By matching the User-Agent header directly in the config, we can now effectively return a 403 response for suspected bots without the overhead of additional plugins. This streamlines our infrastructure configuration and reduces unnecessary complexity.

Updated the Dockerfile to correct a typo in the path for the nobots plugin module when building the custom Caddy image. This ensures the correct package is fetched and built by xcaddy.

Added the workflow_dispatch trigger to the Docker publish GitHub Actions workflow. This allows maintainers to manually kick off a build and push job from the GitHub Actions UI without waiting for a scheduled run or a push to the main branch.

We've implemented a daily scheduled build that runs at midnight GMT+8 for our custom Caddy image. To prevent unnecessary registry updates, the pipeline now verifies if the /bin/caddy binary has actually changed by comparing SHA-256 hashes against the current :latest image before pushing. This keeps our GitHub Container Registry clean and ensures we only trigger releases when meaningful updates occur.

Added a multi-stage Dockerfile to build a custom Caddy binary using xcaddy, including the Cloudflare DNS plugin and full Layer 4 routing capabilities. This is paired with a new GitHub Actions workflow that automatically builds and pushes the image to GitHub Container Registry, making it much easier to deploy custom edge configurations.

This commit initializes the repository by creating the README.md file. It marks the starting point for the Caddy project development.