In the default configuration, killing a vterm buffer with an idle shell will query the user whether they really wants to kill the process. An integration like kitty would account for the shell process when deciding whether killing a session should be silent. The following achieves a similar result for vterm's shell integration.

Since my emacs starts as user service, failing to kill the buffer due to the above will stall a power-off. Since there is no possibility to confirm the query, the user is forced to wait for the systemd user service to timeout on stopping. This will be an annoyance when you're not careful with the default setting.

The solution is simple: emacs has (set-)process-query-on-exit-flag, which disables the query when set to nil. We can toggle the flag when a prompt is printed or a command starts, using OSC 51 message passing. This is done using an escape sequence, cf. the use cases of vterm_printf on the GitHub page of emacs-libvterm. See also the standard XTerm control sequences and Wikipedia article on ANSI escape codes.

During emacs initialisation/using custom.el, we add the following to the list variable vterm-eval-cmds:

("set-query-on-exit" (lambda (flag)
  (set-process-query-on-exit-flag
    (get-buffer-process (current-buffer))
    (not (string= "nil" flag)))))

This allows the escape sequence ^[]51;E<message>^G to evaluate the above designated elisp expression through vterm's message passing function. We also add the following to the shell integration script (e.g. for zsh):

vterm_query_on_exit_on(){
  vterm_printf "51;Eset-query-on-exit t"
}
vterm_query_on_exit_off(){
  vterm_printf "51;Eset-query-on-exit nil"
}
set -A precmd_functions vterm_query_on_exit_off $precmd_functions
set -A preexec_functions vterm_query_on_exit_on $preexec_functions