Design the human-in-the-loop approval gate
Lesson 2 of 5 · Workflows & automation
1 · Learn the move · Constraints + negative instructions
Mark every step where a human must approve before it fires. Anything sent in your executive's name - a reply, an acceptance, a decline, a commitment of any size - never goes unreviewed. In a Cowork routine that means the skill stops and waits for your yes. The gate is the difference between leverage and an apology you have to draft next.
You are a workflow designer. Take my correspondence workflow and add a human approval gate: the model may draft, but nothing sends in my executive's name until I approve. List exactly which steps must never auto-send and why. [workflow]
2 · Your turn. You write the prompt
Your drafted replies could go out over your executive's signature unread. Write a prompt that inserts an approval checkpoint so nothing carrying their name leaves the office until you've seen it.
Write the prompt for this scenario as given, or swap in a real one from your own work. The score works the same either way.
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.