Note to reviewers: This is slightly fragile because we'll only emit hitlStopped if the underlying integration emits it. Some integrations, however (ie Botpress's own hitl integration) might not emit this event when the stop intent comes from a call to the stopHitl action. For now, I believe it's good enough, but if we have a specific use-case that is not working properly, we'll want to either:

• modify every single call to setHitlInactive() to also emit hitlStopped
  • store a conversation tag hitlStoppedEmitted so that we don't emit it twice
• modify Botpress's hitl integration to emit hitlStopped when stopHitl is called (this involves a 3-line change in Skynet)

Note to reviewers: when emitting the events, we also attach the upstream conversation

question: How will we prevent these events from being swallowed by the plugin themselves?

The plugin listens for specifically for hitl:hitlAssigned, which should prevent it from consuming its own event (hitlAssigned without the interface prefix). However, now that you mention it, I just noticed that we're doing beforeIncomingEvent('*') and blanket-swallowing everything 😟. I guess I could modify the handler in hooks.beforeIncomingEvent.all.handleEvent so that it only swallows events if they have a prefix

@franklevasseur see commit 1318d11. The wildcard event handler now ignores events defined by the plugin itself.