Multiple account support in op run

I’d probably keep the account boundary outside the op:// URI and make it explicit at the process boundary instead.

If both accounts need to feed one final command, the least confusing workaround I’ve found is: run op run --account ... once per account, write just the vars you need into separate temp env files, then source or merge those before the real command. It’s a little more verbose than nesting op run, but it’s much easier to debug when one account injects cleanly and the other doesn’t.

So I like the feature request, but in the meantime I’d optimize for explicit account scoping over smarter URI parsing.