I'm parked on you

Pick an option on each card and press Done; your answer goes straight back to the session. The raiser's pick is marked. "I'm not sure" and "Take it to the engineering bridge" are real answers.

1

In plain terms. Should Gera click merge on a one-word fix to Smith's own rulebook, or rule that Smith may land that class of fix by itself from now on? Every test passes and the reviewer says it should land, but the rulebook's own rule says only a person may approve a change to it, so it has sat finished for four and a half hours — and there is already another one queued behind it.
The full question, as raised

PR agent-smith#1139 deletes one stale word from docs/CONSTITUTION.md, Smith's own quality floor. Four of four checks pass at a40bb8cd, zero unresolved review threads, and the reviewer's verdict says 'the one-word edit is correct and I'd land it as written'. Section 11 of that same file says it changes only by a PR reviewed by a human, never by Smith editing its own floor — so no agent has merged it, and it has been finished and green for four and a half hours. Will Gera merge it, and does that rule also cover editorial edits like this one, where nothing about a rule changes?

Why, and what was tried

ASKED BY THE DRIVER of dock smith-alerts-noise-diet. Every figure below was read live from GitHub by this Driver, not relayed from the lane that authored the change. WHAT THE CHANGE IS. PR PropFlow-Technologies/agent-smith#1139, one file, one revision (a40bb8cd), title "docs(constitution): §1's daemon example drops its count". It DELETES the word "ten" from one sentence of docs/CONSTITUTION.md §1 and changes nothing else in the repo. WHY DELETE RATHER THAN WRITE "NINE". The sentence illustrates §1 (DRY) by noting that a derived daemon list already existed when a proposal nearly added a hardcoded second one. The count was never the load-bearing part; the clause carrying the argument is "its list derived from the propflowai manifest via daemon_labels()". And the dates say the numeral was never a miscount: on 2026-07-19 the probe really did cover ten (eight daemons + zoomhook + the not-yet-retired marketing-thread-editor tombstone). It went stale one to three days later and nothing noticed for three months. So "nine" would be equally correct today and equally unguarded tomorrow. Writing a corrected numeral resets the clock; deleting it ends the class. THE MECHANICAL STATE, read live. Four checks, all SUCCESS at a40bb8cd: gate, lint-test, "plan review (Astra)", review. Zero unresolved review threads. The reviewer's verdict is 🟡 and its first line is "the one-word edit is correct and I'd land it as written. The only finding is in the PR body, not the diff. Per §11, Gera decides whether this merges; this verdict doesn't." THAT SINGLE FINDING IS ALREADY ADDRESSED, and the Driver verified it at source rather than taking the lane's word. The verdict said the body wrongly claimed the bullet's "one daemon that was already retired" referred to marketing-thread-editor. The body now carries an explicit section headed "Correction to an earlier version of this body", removes the claim, gives a dated table with receipts for each row, and deliberately leaves unsettled WHICH daemon the original sentence meant rather than adopting the reviewer's own unverified guess. It needed no new revision: measured across every file in .github/workflows/, no workflow here lists the `edited` event, so a body rewrite fires nothing. That is also why the glyph is still 🟡 and cannot become 🟢 without a push nobody needs. THE QUESTION, AND WHY IT IS NOT MINE TO SETTLE. docs/CONSTITUTION.md §11 (:209) says the document "changes by PR, reviewed by a human (Gera decides contested rules), never by Smith unilaterally editing its own floor." A bot verdict is not a human review. The PR was authored on that basis: its body states up front that it is waiting on a person and not on a check, and asks that nobody ✅ it into the auto-merge queue. So this Driver has deliberately not merged it, and is not asking for permission to re-read the rule in its own favour — an agent deciding that the clause restricting agents does not bind it is the exact shape §11 exists to refuse. WHAT IS ACTUALLY BEING ASKED is whether the right next act is to put it in front of Gera now. The work has been finished and green since about 2026-10-16T00:05Z and has sat untouched for roughly four and a half hours because nothing in the normal path routes a §11 change to a person. If the answer is that it needs him, this is the routing request. If there is an existing standing authorisation that covers a numeral deletion in that file, naming it would close this question for every future row of the same shape, of which this dock has one more already filed. THE COST OF LEAVING IT. It is one word in a document used as the quality floor for every review in the repo, so nothing breaks while it waits. But the stale numeral is currently quoted as fact by anyone reading §1, and the dock row behind it cannot close, which keeps a finished item open on the board Gera reads.

The session raising this recommends: I'll merge it — only a person may change Smith's rulebook
Fable RESOLVED this rather than escalating it, and its resolution was 'Route it to Gera now for the merge' (receipt f0b4598a6, two passes with the options reversed, agreeing at high confidence). So this is not an undecided question being passed up — it is a decided one being carried out. What is needed is an ACT, not a judgement: section 11 reserves the merge click itself to a person. Fable's reasoning, checkable at the receipt: that section carries no de-minimis carve-out; 1,443 ledger entries hold no ruling about Smith's constitution; no ADR in either repo touches it; and the file's only two revisions are both Gera's, so every prior change was human-reviewed. The Driver's standing merge authority is generic and risk-based, and a specific written rule about this one file beats it. Fable also asked that Gera rule on the CLASS, because this dock has a second same-shape row behind this one and only a section-11-level decision can mint that carve-out. Cost if wrong: a few minutes of his attention on a one-word docs edit. The asymmetric risk runs the other way — an agent merging its own quality floor, however trivially, is exactly the precedent that section was written to refuse.
Tried first, unsuccessfully: fable-decide was run BEFORE this raise and RESOLVED it (receipt f0b4598a6, two passes with the options reversed, agreeing at high confidence). Its answer was 'Route it to Gera now for the merge', so nothing is being escalated past a decider that could have settled it — it settled it, and what it settled was that the remaining step is a person's click. Fable searched for a standing authorisation that would let an agent merge this and found none, naming each place it looked. It could not reach gh from its seat, so the mechanical state (4/4 SUCCESS at a40bb8cd, zero unresolved threads) is as this Driver read it live; Fable noted its choice does not depend on it.

ff9239cf is parked on this.

2

In plain terms. Should Gera reopen a finished, approved change that was closed by accident one second after an unrelated change merged? It is five files of redaction work with a green review on its current version, and nothing is lost yet — but an automated cleanup step has already asked to delete its working copy, and this Driver is not allowed to reopen another team's work.
The full question, as raised

PR agent-smith#1102 was closed WITHOUT MERGING at 2026-10-16T06:54:21Z, one second after an unrelated PR (#1147) merged at 06:54:20Z. The evidence says it was not intentional: mergedAt is null, the close event's actor is the shared gera-propflow token with commit_id null (GitHub's keyword auto-close populates commit_id, so it was not that), #1147's merged commit carries NO closing keyword for it — the three #1102 mentions are all prose in a paragraph naming it as the place to observe the fix working — and nothing else closed in that window, so it was not a sweep. This is the second recorded instance: #965 closed unmerged 3 seconds after #968 merged, same shape. At stake: 5 files, +655/-96, 48 comments, and the newest reviewer verdict on it is a GREEN LGTM at 70c7ad21, its own current head. Nothing is lost yet — the branch is still on origin and its worktree is intact — but a post-merge hook already told this Driver to tear that worktree down as 'finished', which was refused. Reopening it is not this Driver's to do: it belongs to no row on this dock, and the row fence refused the write for exactly that reason. Who reopens it?

Why, and what was tried
The session raising this recommends: Gera reopens it and the work carries on
This Driver cannot act and should not: #1102 names no row on this dock, it belongs to different work (a standup-recap redaction change, authored by the agent-smith app on 2026-10-15), and the row fence refused with reason item_unnamed — its own warning being that fencing an effect against a guessed row is how an effect lands on somebody else's work. Reopening under one of this dock's row ids would be precisely that. No live session on the roster (48 listed) carries a name or tmux suggesting this work, so there is no peer Driver to hand it to. The asymmetry decides it: reopening is a one-click reversible act, while the worktree teardown the closure triggers is not, and a hook has already asked for that teardown once. If the closure was in fact deliberate, closing it again with the reason stated costs one click and this card becomes the record of why it looked like an accident. Option 2 is refused by the fence and would be this Driver widening its own authority over another dock's work. Option 3 discards a reviewed GREEN change by default rather than by decision — and closing a PR discards the parts nobody reassigned, which this fleet has already paid for once (six wrong-channel fixes vanished with #1095 until a reviewer found them two parts later).
Tried first, unsuccessfully: fable-decide was NOT run before this raise, and the reason is the clock rather than the question: this is a one-click reversible act on somebody else's reviewed work, a hook has already requested the destructive step that follows from the closure, and the last two decider rounds tonight each took Fable's as-given pass past a 600-second timeout and fell through to Astra at medium confidence — so a third round would cost ten minutes against an action whose cost if wrong is one click to close it again. Everything in this card was measured by this Driver directly rather than inferred: the close event's actor and null commit_id from the issue timeline, the absence of a closing keyword from the merged commit message, the 🟢 LGTM and its sha from the newest glyph-carrying reviewer comment, the branch still present on origin at 70c7ad21, and the worktree still on disk. The only thing not measured is WHY it closed, which is the part a model cannot settle from here.

ff9239cf is parked on this.

3

In plain terms. How long should a problem have to stay fixed before Smith announces that it recovered? Right now, when something breaks, gets better, and breaks again, Smith announces it every single time -- so the things that flap on and off are what fill Gera's DM. Waiting 30 minutes before saying 'recovered' would cut it from about 103 messages a day to about 56, and Gera would still always hear about anything new that breaks. The quietest option goes further but can stay silent about a fault that keeps coming back.
The full question, as raised

Your DM takes about a hundred automated posts a day. The biggest single cause is that a problem which clears and comes back is the LOUDEST thing we can send you, not the quietest: host-alerts dedups one post per SPELL of a condition per cooldownS (4h default), so the quiet period only applies while a problem STAYS broken. Anything that flaps starts a new spell and re-pages immediately. fleet-medic has the same shape -- alarm_repost_s is 21600 but its RESOLVED loop deletes the record the quiet is keyed on. One stale fleet-medic condition flapped 19 times in 23h. All four numbers below are a replay over the same measured window.

Why, and what was tried
The session raising this recommends: Wait 30 minutes clear before announcing recovery (about 56 a day)
It nearly halves the volume without ever silencing a NEW break; only the fourth option can make a real intermittent fault silent, and it also breaks 13 existing checks.
Model receiptFABLE was asked, and escalated. Receipt f8bc8f1a2 — NEEDS HUMAN.
The threshold — how long a condition must stay clear before 'recovered' is announced — sets how many posts his own DM takes versus how soon he hears a flap settled. That is attention/risk tolerance. No ruling on disk fixes the number: roster_alert's 600s is an engineering default on a 15-min-cooldown pager, and prior rulings (b90701615, b90854896) support cutting non-actionable noise but never pick a duration. Raise it with the four priced options and the lean to 30 minutes stated.

(both passes said so.)

ff9239cf is parked on this.

Answered and withdrawn (8)
Pick an option above, then press Done.
PropFlow Docs