It’s true that Claude Code will continue to suggest further things to do, seemingly forever. But I’ve never had an issue making the call of “this is good enough” or “that’s not necessary right now”. If you have a clear idea of what you want to accomplish, it’s easy to get a task 100% completely done with just a few prompts. There’s only a paradox if you don’t have a clear end state in mind and let the agent continually move the goal post of what “done” means.

There's a crass Zeno's Paradox joke involving an engineer, the devil, a pretty lady and a football field: the punch line being something like "I know the paradox but can get close enough to do what I got to do."

Regarding TODOs in code I really like the philosophy of including a ticket with each todo item and pasting the ticket url in the todo for follow up and accountability... YMMV