What truly needs them
Lesson 2 of 6 · The Inbox
1 · Learn the move · Constraints + negative instructions
Most assistants over-flag, and it isn't laziness, it's fear of being the person who sat on the one that mattered. Positive rules don't fix it, because everything looks important at 7am. Negative rules do: write down what NEVER gets flagged. No FYIs. No meeting confirmations. Nothing you're already authorized to answer. Nothing a "got it" would close. Hand the AI that ban list and it stops asking your executive to bless your inbox. What survives the bans is the real flag pile, and it's short.
You are helping me write my flag rules. Output four sections: NEVER FLAG (at least ten bans, one reason each, built from the authority I list below), ALWAYS FLAG (the short list), THE GRAY BAND (the cases the bans don't settle, each as a question with a default), ESCALATION OVERRIDE (what beats every ban). Build only from what I paste. Holes get [NEEDS: confirm with your executive], never your choice. [my authority, my named senders, my limits]
2 · Your turn. You write the prompt
You flagged 19 messages Monday. Elena answered four and forwarded the rest back with "you can handle these." Write a prompt that turns the flag decision into a written rule set built out of what never gets flagged.
Remember: the AI sees only your prompt, not this page. If the situation isn't in your prompt, it doesn't exist.
Optional. These shape the output when you run your prompt below, not your score.