CLI anatomy
Every Elegant Git command line has the same shape:
[binary] [object] [action] [argument]... [modifier]...
Knowing the five parts is enough to read any command in these docs, to guess a command you have not seen yet, and to script the tool without surprises. So, one part at a time.
Binary
The binary is called eg. After eg self configure, you can also run it as git elegant …: configure writes the git alias alias.elegant = "!eg", so Elegant Git sits next to the rest of your Git commands.
Two commands sit on the binary itself, without an object, because they are about the program rather than about anything it manages:
eg version
eg completion bash
version prints the installed version. completion writes a completion script to stdout for bash, zsh, fish, or powershell. The shell name is required; in interactive mode it is asked if you leave it out.
Install bash completion like this:
eg completion bash > ~/.local/share/bash-completion/completions/eg
git elegant <TAB> does not complete (the git alias is a ! expansion). Use eg <TAB>. If you previously installed a git-elegant completion file, delete it.
The hidden legacy flat names — start-work and its siblings — also live at the binary level, as deprecated aliases.
Object
An object is the thing you are talking to. There are six of them: self, workspace, repo, work, release, and hook.
Action
An action is what you do to an object. Together they are named {object}.{action} — for example work.start. That dotted form is what you pass to hook new and what appears in per-command state, so it is worth recognizing.
Running eg with no object prints help outside a work tree, in a repository with no commits, and when prompts are off (--non-interactive, CI, or a non-TTY). Inside a repository with prompts on, it runs the detected work action. When detection does not choose an action, or that action still leaves a choice, it opens a picker of the commands that fit that repository and runs the one you choose.
Running eg <object> with no action prints the action list, except for work and workspace: in interactive mode those two detect the next action from context, or ask you when more than one fits. In non-interactive mode an action is always required. See interaction.
Argument
An argument is a positional input to an action, required or optional. An action can have anywhere from zero to several of them, and they are all resolved before the action starts to work. If something is an input to the action, it is positional — you will not see a flag for it. How missing arguments are asked, or refused, is on the interaction page.
Modifier
A modifier changes how an action runs without changing what it does. There are three, and they are global: --no-workflows, --non-interactive, and --interactive. They may appear anywhere on the command line, so eg --no-workflows work save and eg work save --no-workflows are equivalent.
--no-workflows— skip ahead and after hooks--non-interactive— no prompts; fail when a required input is missing (alsoELEGANT_GIT_NON_INTERACTIVE=1,CI, or non-TTY stdin)--interactive— on a TTY, force prompts (overridesCIand--non-interactive; alsoELEGANT_GIT_INTERACTIVE=1). Ignored when stdin is not a TTY
--version prints the version, same as eg version. What each interaction modifier means is on the interaction page.