I was intuitively doing some of these (e.g. providing verification criteria, limiting the scope of each prompt, specific tasks rather than vague ones), but others were somewhat new to me (e.g. continue / resume session, frequent "clear" to remove junk from the context).
Useful reference.