The situation
Something breaks. It gets fixed in a chat thread, and a week later nobody can say how many times it happened, or which product has been quietly failing all month. The health of the portfolio is whatever the last person remembers.
1. Open the incident where the product is
create_incident against the product, with a severity. That is the whole ceremony - it is meant to be cheap enough to do at 2am.
2. Let health follow from it
The health band is COMPUTED, never typed in: unresolved sev1 or sev2 puts a product in Critical, sev3 or sev4 in Watch. get_product_health explains the band rather than asserting it - every number comes back with what produced it.
That is why nobody has to maintain a status field. A product is critical because something is open against it, and it stops being critical when that is resolved.
3. Close the loop
update_incident when it is resolved. The stamps are the record: opened and resolved are facts with times, not a status somebody has to remember to change.
4. Ask the agent for the morning read
Call get_workspace_overview. Tell me which products are Critical or Watch and WHY - name the incidents. Then check list_changes since yesterday for anything I have not seen.
5. Write down what you learned
An incident that taught you something is a decision waiting to happen: what you chose to change, what the alternatives were, how confident you are. Link it to the incident so the next person finds the reasoning next to the event.
What this buys you
The portfolio view stops being a guess. "Which products are in trouble" has an answer that follows from records somebody made at the time, and every band on the screen can be expanded into the incidents that caused it - which is the difference between a dashboard and a rumour.