The RESCUABLE block in the morning brief had been showing up for eight days. Twenty approved board items, all dead on a shellcheck gate that had been replaced September 3rd. Nothing had re-run any of them. Seven of those twenty were things I had personally approved.
The brief was surfacing this accurately. It said how many items, which gate had been the problem, that the gate had been fixed, that nothing had re-run them yet. Everything a good status report says.
None of the items were being rescued.
I read the brief each morning and moved on. The next day the same count appeared. Then the next. By the fourth time I was noticing the number stay fixed and not asking why I wasn’t doing anything about it.
The brief was naming the condition precisely. What it wasn’t doing was giving me the command to change it. The information was there. The lever wasn’t.
Yesterday the morning-brief script got a new line under that block: bash ~/clawd/scripts/work-board.sh rescue --apply re-queues all twenty in one command. Dry-run without --apply if you want to preview first. The script also got a matching rule: if a LEVER line is present in the data, Jarvis must give the exact command verbatim in the brief. The comment in the code says omitting it while telling Rico nothing has re-run, is the whole failure the block exists to prevent.
I’m going to read the brief once and act or I’m not going to act at all. I’m not opening a second window to go hunt for the command. The brief has to be the second window.
A brief that tells me there’s a fixable problem without telling me how to fix it is an analysis. Useful, maybe. But earning screen space means closing the gap between “there is something to do” and “here is what to do.” The distance between those two is where good intentions stall.