Could this be re-purposed for detecting anomalies/outliers in time series data?

If you define anomaly as something unexpected then yes. In this case, if the reality differs significantly from the forecast (=expectation) then it is an anomaly (according to our definition). In numeric univariate case, there could be positive anomalies where you get more than expected, and negative anomaly where you get less than expected.