Authorship fights start as assumptions nobody wrote down. Two people each believe the list is obvious. Neither says it. Eight months later the paper is late and the list is a wound.
This is not legal advice, not a journal policy, and not a substitute for your research office. It is the message that puts a proposed list in front of the group while changing it still costs an email.
Propose a list, marked as a proposal
Write the names in the order you are proposing, and one line each on why they are there: designed the study, ran the cohort, wrote the first draft, supervised. If you are guessing about someone's contribution, state that assumption. "I assume Jordan wants corresponding-author duties" is a sentence they can correct. A tidy list your assistant inferred from a chat is how you insult someone who did the work.
The collaboration request already says to raise credit while it is cheap. The group version is the same move with more names. Invite edits. "This is a starting proposal, not a carved tablet."
Do not settle grant language or data-use agreements in the same briefing. Those belong with people who are paid to read them.
Ask for accept, edit, or meet
Ask for a decision: accept this list and order, send a written edit, or meet because the list is not closable on the page. That is a response shape with an owner. Accept means you will submit with these names. Edit means they return a different list. Meet means you stop pretending writing is enough.
A sufficient reply is short. "Accept." "Edit: add Sam after the methods line." "Meet — I think the order is wrong." You did not ask for a history of the lab.
Give a date and a default you will actually use. "If I do not hear by the 10th I will send the present list to the corresponding author as the working draft." A decorative date is how the hallway rumour becomes the list.
Keep the record, and keep the model honest
You draft. You read every name. You confirm the send. Credit under your name is not a paragraph the model gets to mail because it sounded fair.
People without accounts get email and a link. People who already have accounts need an accepted contact.
No attachments. The draft manuscript stays in the shared folder. The briefing is the list and the ask.
Nicknames stay private unless you share them. A private word for a labmate does not belong on a paper conversation they will re-read in a year.
If someone blocks you, the block is silent. That is not a verdict you get to narrate. Take the formal path your office already has.
When the proposed list is short enough to accept or edit in a line, connect your assistant and ask while the work is still small.