We don't need to limit ourselves to existing interfaces anymore: with agents it's also possible to have them build a tremendous amount of scaffolding to support comprehension and make changes in ways that are digestible.
Often enough for larger, complex stacks I have the agents build me a custom view for the Pull Request with 4 columns: the standard 2 for the diff, a 3rd column with running commentary from the agent on why it made a particular change and finally the 4th column as a place for me to add comments and directions.
Built it for PR #5, the deploy change that keeps local edits to tracked files (the .htaccess stash/pull/pop case):
https://claude.ai/artifact/AHrjEnnZGyKyH9mBdUTtBG
Before, after, my commentary, and a notes box for each hunk. Notes stay in your browser, and "Copy notes" bundles them with file and hunk references so you can paste them straight back to an agent as directions.
One addition: the files are in reading order instead of alphabetical. The deploy routine comes first, then the tests, then the template, then docs. Where something deserves a closer look, the commentary says so, for example the single reset --hard and the local_changes reason code that no longer exists.
I'd make the commentary column collapsible per hunk. For the deploy example, read the diff first and write down what you expect to happen if the pull succeeds but the stash pop conflicts. Then reveal the agent's explanation. Any disagreement gives you a concrete test to run before approving. You only need to do this on the few hunks where a wrong guess would matter.
