Everything I could decide myself is decided. What's left needs
you. Each one carries the recommendation of the session that raised it — if you
agree, pick it and press Done. A model's opinion, where there is one, appears on the card
as a receipt line; cards with no receipt were never put to a model. "I'm not sure" is a real answer and becomes work for me.
You don't need to open the session — pressing Done sends your answer back and it picks up.
1Task appfolio-pool-probe: operator-here refuses to arm the launch acceptance.json (items 1-3 are 'file:' checks; agent-smith #467 made file:/exit: ATTESTED and unpassable; only pr:/run:/slack: are observed). A Slack-witnessed re-cut (POOL-PROBE-PREFLIGHT / POOL-PROBE-SAMPLES / POOL-PROBE-READOUT tags in C0C10AS4U15/1789505698.307749, same substance, preflight still posted BEFORE the run) is refused exit 4 by the monotonicity invariant: rewriting an item routes through the human door. Re-baseline needs a CHANGED criterion text (identical text = identical hash = still refused). On 'Re-baseline' I run: drive --task appfolio-pool-probe --replace "<criterion + one sentence naming the three tags>" then operator-here --criterion-file <new> --acceptance-file /private/tmp/claude-501/-Users-miniclaw-code-PropFlow-propflowai/b53e3d6d-8550-45eb-baca-af095cbda351/scratchpad/lanes/_driver/acceptance-recut.json, paste the ARMED read-back, then spawn the Worker. Alternatively run it yourself on the mini: the new criterion text is at /private/tmp/claude-501/-Users-miniclaw-code-PropFlow-propflowai/b53e3d6d-8550-45eb-baca-af095cbda351/scratchpad/lanes/_driver/criterion-new.txt.
In plain terms. The AppFolio pool-probe task was launched with a finish line that can never close (three of its four checks look for files the driver itself writes, and the watcher refuses those as self-attested), so nothing is watching this task and the probe cannot start — may I re-baseline the finish line to the same three receipts posted in the bridge thread, where you can read them? Answering 'Re-baseline' lets me run the one-line re-baseline, re-arm, and start the probe; 'Keep as launched' leaves the task unarmed and nothing runs.
The session raising this recommends: Re-baseline to Slack-witnessed receipts Same three receipts, harder witness: a thread humans read instead of a file the driver writes; the invariant only requires that a human authorise the change. Tried first, unsuccessfully: receipt f91357e40 — RESOLVED by ASTRA (gpt-6-astra); Fable was unreachable (weekly limit, resets Sep 17 19:00 CT). Astra's ruling: 'A human runs drive --replace — the Driver reports verbatim to the bridge thread and decisions dock with the exact command and re-cut list, and waits.' It is not a NEEDS HUMAN verdict, so this is raised on free text: the ruling names a human as the required actor, and this block is how that human hears it and how the answer routes back to the raiser to execute.