obvi | Polymer 3 webcomponent / button for doing speech | Speech library

I need to run a GUI app on the logon screen on Windows 7 to 10.

I have a nodejs app registered as a service which runs as SYSTEM (the default), using node-windows module. This seems to be a prerequisite to what I need, but something more is necessary.

The app is just a basic socket client that connects to a socket server, nothing special as of yet.

I am thinking maybe if I use child_process.spawn() to open an elevated command-prompt and/or runas another user-- but what user? SYSTEM doesn't do it. I'm thinking whatever winlogon.exe runs as, but that doesn't seem to be SYSTEM even though that's what it shows as in task manager.

Is there a windows command to run a GUI in the logon screen if you have SYSTEM-level access?

UPDATE:

Ok. As of now I have a reverse telnet-like client running on the "remote host" as a Windows Service under the SYSTEM user. It periodically connects to a test server text interface (also node), from which you can enter nodejs commands-- basically a home-brew JS REPL.

I am able to spawn a hidden, interactive windows commands prompt under Session 0, type commands and get output through my little reverse command-line terminal.

Obvi, encryption and security will be important here, but for now Windows is running in a VM inside Linux on my laptop, and the app(s) only talk to each other through private NIC's on a Host-only Virtual Network, so packets are not spilling out onto the the real network.

At my windows command prompt in Session 0, I tried running my remote-control GUI app. Instead of being presented with a GUI, I get the following error:

[5472:5492:0828/031356.901:ERROR:gpu_process_transport_factory.cc(1024)] Lost UI shared context.

If I tasklist, I get:

winlogon.exe 2140 Console 3 6,704 K Unknown NT AUTHORITY\SYSTEM 0:00:00 N/A
LogonUI.exe 4772 Console 3 19,104 K Unknown NT AUTHORITY\SYSTEM 0:00:00 N/A

So, I can see winlogon.exe and LogonUI.exe are running in session 3 as user 'SYSTEM'.

How can I put the service in the GUI context of winlogon and logonui, or spawn my app in it's ?

Is there a windows commands to do this? A nodejs command/module/function?

MORE INFO:

It needs to be a programmatic, automatable solution I can work into the program-- no manual tweaking using another GUI app to get it working.

Registry edits or anything accomplishable via node or windows console commands should be acceptable, however, since those I can incorporate into my code.

Background: This is mostly to run a remote-desktop app I wrote, much like Screen-Connect or LogMeIn. It works great-- except for the part where it pauses anytime you invoke anything calling up the UAC or enter the logon screen, making it not yet practical for remote IT work.

The nodejs windows service app also works good for it's part, autostarts as a service at boot time, autoconnects/reconnects to the test server without requiring login or a continuous connection, and stays alive persistently.

Basically I am thinking if I can run an instance of it in the logon screen it should be able to then grab desktop video capture for output to the network, and invoke remote mouse and keyboard input from the network.

The desktop-capture is done using NW.js, and child_process.spawn()ed robotjs-based socket-client running on the remote host.

Unfortunately, implanting robotjs in the service app doesn't work-- I tried that as I was thinking at the very least I could have a custom remote logon screen in the controller-app and have robotjs input the password in the remote pc's logon screen-- but no luck. It appears I am only getting the legacy Desktop 0, not the logon screen for robotjs-based keyboard/mouse input.

I am really hoping all I need is to 'runas' some other user or process. **Note: I am looking for a way to do this that does not involve auto-logon, reboots, or workarounds. I know MicroSoft has officially supported ways for accomplishing such tasks-- but I have yet to find a way to do it through Node rather than C. If there's a node-module using native code to do it, that's fine by me. I just want my code to be JavaScript.

Anyway, that is the actual goal I'm asking this question for-- to run my remote desktop app on the logon screen to grap video and invoke mouse actions and keystrokes.

If there is a better way that doesn't match the question in the title, please do answer and I will edit the question to match the answer, if it works.