Whether the agent sees the capture now: it is on, the channel lets the agent see pixels, and the model is being sent its frames (its source among the session's video sources is on and active, which is what the "Agent can see" chip reads). On a model that takes one video stream, a screen share started after the camera is the one sent, so the camera is not seen until the user picks it again.
The shared panel's name, while the user shares one panel of the page; null otherwise.
The user's camera as mj-self-view takes it: while the camera is on, the same object for one stream.
Why the last start failed, while it stands.
The agent's reason for asking, while its request stands.
The shared surface as mj-share-preview takes it: while a share is on, the same object for one stream.
What a share shows.
What the capture surface shows.