Vision check-in

This is the controller that closes the loop between the two worker skills. find-risky-assumptions surfaces the bets a vision makes and owns risk (stakes). derisk-assumptions runs spikes and owns confidence (signed belief). This skill owns neither axis — it reads the current state of the ledger, tells the user where the vision stands, and drives the next human decision. Many spikes may be in flight at once (multiple agents de-risking in parallel), so a check-in is always a snapshot: report what is settled, name what is still running, and act on what needs a human.

What a check-in decides

Run the classifier, then have exactly one of these conversations (a real ledger often has ingredients of several — lead with the highest-priority one):