I was also just looking for this.

My scenario is that I need to create aggregate datasets (CSV or Parquet) and then push them out to S3 (and invalidate a Cloudfront cache), where they are read by a dashboard.

Bonus points is there is some kind of template variables that I can use in the output filename (date formatting, random UUID etc).

Yu, great idea as for now I do this on third party tools like Kaggle or GitHub

This would allow me to replace a huge portion of my flat file transformation processes and would be a huge win for me, architecturally.