EXEC %|[-] [-nosh] [-out | -msg <target> | -notice <target>] [-name <name>] <cmd line>
EXEC %|-out | -window | -msg <target> | -notice <target> | -close | -<signal> <id>
EXEC %|-in <id> <text to send to process>

     -: Don't print "process terminated ..." message
     -nosh: Don't start command through /bin/sh
     -out: Send output to active channel/query
     -msg: Send output to specified nick/channel
     -notice: Send output to specified nick/channel as notices
     -name: Name the process so it could be accessed easier

     -window: Move the output of specified process to active window
     -close: Forcibly close (or "forget") a process that doesn't die.
             This only removes all information from irssi concerning the
              process, it doesn't send SIGKILL or anything the process.
     -<signal>: Send a signal to process. <signal> can be either numeric
                or one of the few most common ones (hup, term, kill, ...)

     -in: Send text to standard input of the specified process
     -interactive: Creates a query-like window item. Text written to it is
                   sent to executed process, like /EXEC -in.

Execute specified command in background. Output of process is printed to active window by default, but can be also sent as messages or notices to specified nick or channel.

Processes can be accessed either by their ID or name if you named it. Process identifier must always begin with '%%' character, like %%0 or %%name.

Once the process is started, it's output can still be redirected elsewhere with the -window, -msg, etc. options. You can send text to standard input of the process with -in option.

-close option shouldn't probably be used if there's a better way to kill the process. It is meant to remove the processes that don't die even with SIGKILL. This option just closes the pipes used to communicate with the process and frees all memory it used.

EXEC without any arguments displays the list of started processes.