When a command fails during interactive compute usage, the error message is constructed using the results of the command (exit code, stderr). However, it would be useful to ensure this is available directly on the error object rather than providing only the resultant message.