The Agent Skills Directory

[COMMAND_EXECUTION]: The script executes docker and curl commands. It properly uses shell array expansion ("${build_args[@]}") and input validation to prevent command injection. App names and tags are strictly validated against allow-list regular expressions.
[CREDENTIALS_SAFE]: The skill handles registry credentials (ROBOT_USER, ROBOT_SECRET) using secure patterns. It accepts them via environment variables or protected configuration files, uses docker login --password-stdin to avoid leaking secrets in process lists, and utilizes a temporary curlrc file for API calls to keep credentials out of command arguments.
[DYNAMIC_EXECUTION]: The script includes a custom configuration parser (load_config_file) that manually parses KEY=value pairs instead of using the shell's source command. This effectively prevents arbitrary code execution (RCE) that could occur if a malicious configuration file were loaded.
[SAFE]: The skill implements operational safety features, such as a 'dry-run' mode and a network latency check (measuring TLS handshake time) to prevent extremely slow cross-region image pushes that could hang deployment pipelines.

xgent-image-push