A "scenario" is the general term we use to refer to a unit of work the agent needs to perform, monitor or manage. There are four scenario types: run, capture, watch and unix socket.
The agent will launch the given command on either a set frequency, e.g. every 60 seconds, or at a scheduled time, e.g. 15 minutes past the hour. As the command runs, the output (and errors) will be captured and sent to the cloud.
The agent will monitor the given log file and send any new entries to the cloud.
The agent will watch for any changes to the target file and generate an alert. If required, the agent can also send a copy of the updated file. Perfect for monitoring configuration files.
This will allow you to have long running processes that can talk to the cloud via the agent without the need for an intermediary file. Please see the article on Unix Sockets for more details.