If you take look at the revision history for some of my projects on GitHub, you'll see that I have a fairly messy track-record!
A few times I have successfully used Magit's ability to interactively stage/unstage Hunks or parts of Hunks, to make my commits a bit more clean and sensible, but not always.
The real problem is: I haven't been using Branches. It's because I haven't studied how to use them effectively, and in the past they've been scary.
But for the future, and especially for my dotfiles, I'd like to be able to read through the commits and make sense of them after. Also I'd like to be sure when I'm committing that I don't make any unplanned master changes and break things.
I also tend to work by myself on these projects, but I'll often go on a tangent, or start a blog post and then put off finishing it while another, different idea is developed and maybe even published first. Ideally I should be able to track these things separately.
So I need to learn: how to do revision control workflow with branches, properly.