LABELS.md's operator boundary excludes a third party's unscheduled reply — #573 D3 named the exclusion and left the class unlabelled, and la-familia-incubator#20 is its first instance

[dan-claude-bot]

The rule, quoted at this repo's pin

heavy-duty/la-familia-incubator runs ceremony 0.7.8. LABELS.md's operator passage — added by #573, and byte-identical on main today:

An issue waiting on an external condition only the operator can resolve is operator-owned: it keeps ready and adds operator, and its quiet is legitimate under the same seven-day clock. This admits conditions the operator has the power to change, such as rerunning a red fork-PR head; it does not admit a third party's unscheduled reply, which the operator cannot resolve.

The exclusion is deliberate and it is #573 D3 verbatim — "a red fork-PR head that only they can rerun, and not, for example, a third party's unscheduled reply. State the boundary, or the flag becomes the drawer everything unactionable goes into and its clock stops meaning one thing."

The case the rule cannot express

heavy-duty/la-familia-incubator#20 is open on one acceptance criterion, seven of eight ticked and its pull request merged on 2026-08-31:

Operator-owned. Before any deployment serves this form, legal/04-collection-point-es.md is adopted by La Familia's counsel. Evidence: the document's own front matter and body at the adopting commit. Wake condition: counsel returns the adopted text.

The wake is a third party's unscheduled reply, in the exclusion's own words. The document carries owner: La Familia — counsel in its front matter and says so itself: "La Familia's counsel adopts, rewrites or rejects it and owns the result." Heavy Duty Builders Labs prepared the draft; the operator (OPERATOR="${HUMAN_REVIEWER:-danmt}") can chase counsel and cannot return the document adopted. So the criterion is real, tracked, correctly split out under #488's route — and its issue has no admissible label.

This is one instance and it was measured one-directionally: the other four operator issues on that board were checked against the same boundary and all four are admitted — the GHCR Manage Actions access grant (#12 crit 4, #11 crits 1–2), the 0.1.3 cut (#11 crit 5, #12 crit 6), a credentialed workstation (#14 crits 1/2/5) and the bootstrap=yes press (#31 crit 6) are all acts the operator has the power to perform.

Both available shapes are wrong, and the second one is measured

shape	why it fails
ready + operator — what #20 carries today	the boundary above excludes it by name
ready alone	owner_class_decision at 0.7.8 returns ADD_BUILDER for enhancement ready scope:template, and KEEP for the same set with operator — run in one shell with controls (epic → KEEP, plain ready → ADD_BUILDER). That writes builder onto an issue whose one remaining criterion is nobody's build
blocked	needs a parseable declaration and there is no issue to name — #573's own table already ruled this out for the class
needs-ruling	the pending thing is not a decision this fleet's decider owns; a bare flag is noise (#50 D12)

⚠️ The ADD_BUILDER write is scheduled rather than hypothetical. It is inert on that board only because the builder row does not exist there yet, and the reconciler gates the whole derivation on BUILDER_LABEL_AVAILABLE. That row arrives with the same bootstrap=yes press #31's last criterion is waiting on — so the day the operator clears one item on that board's human-owed list, the other shape starts writing a false label.

⚙️ One thing that is not broken, said so the report is one-directional: the seven-day nudge's second limb — "or say what is still outstanding" — is satisfiable by the operator, so the nudge is not the harm. The harm is the classification and the absent alternative.

The options

1. Widen the admission and keep D3's containment in the body. operator also admits an external condition the operator can chase but not resolve, provided the criterion's wake names the third party. One sentence in LABELS.md, no taxonomy row, no sweep branch, reversible.
2. Keep the exclusion and give the class its own shape — a taxonomy row, or a documented ready + prose form exempted from owner_class_decision. Most expressive; #573 priced a row at "nine boards' worth of surface plus a sweep branch" and declined it on one instance.
3. Rule the class off the board — such a criterion is recorded somewhere other than an issue. This gives up the tracking #488 exists for, on the criterion most likely to be forgotten.
4. Rule that counsel's adoption is not the excluded case — that "chase" is enough to make it operator-resolvable. Then D3's exclusion needs a different example, because this is the clearest one the fleet has produced.

Recommendation: 1. D3 wrote the boundary to stop operator becoming the drawer, and that risk is real — but the containment it needs is the body's wake naming who owes the reply, which #491's evidence-surface contract already forces. What the exclusion buys today is not a narrower drawer; it is an issue with no true label at all, whose only alternative asserts a builder owns work no builder can do.

The workaround now in force, and what retires it

⚙️ #20 keeps ready + operator, as the less-wrong of two wrong shapes and the one that keeps builder off it, with the reading written into the criterion's own text and this discussion cited there. Nothing on that board is gated by this — the criterion's wake, its evidence and its owner are unchanged, and no builder is misled today because that board carries zero builder-owned rows.

What retires it: a ruling here. Under option 1 the workaround becomes the rule and the citation on #20 is deleted; under 2 or 3 that issue is relabelled or re-homed in the same tick.

— triage, heavy-duty/la-familia-incubator, 2026-09-07.

[dan-claude-bot]

🧭 Converged — escalate, made in place. The thread stays open as the venue, and the ask is smaller than the body above it

@danmt — this thread is an escalation and you are the decider. The body above named nobody, so since 2026-09-07T02:14:22Z it has notified no one; that omission is mine and this comment is the repair. Nothing is minted until this is answered.

Re-measuring before converging is where a thread I wrote myself stops being a proposal, and here it changed the finding. The body asks "which of four new rules should operator grow?". The measurement says the rule you already ruled probably answers the case, and what is actually broken is one sentence's scope. So the question below is not the body's question.

---

🧭 needs-ruling — does #573 D3's exclusion bound the admission D1 added, or the whole operator class?
Options: A — it bounds the added admission; the base class is untouched B — it bounds the class itself
Recommend: A, because it is the only reading under which LABELS.md and TRIAGE.md's reach axis do not contradict each other.
Blocked: the LABELS.md edit and any mint stop here. Everything else continues — la-familia-incubator#20 keeps ready + operator under either answer, no label moves, no criterion ticks, and that board carries zero builder-owned rows so no builder is misled today.
Default: none — hard block. The chosen reading is the shipped sentence, and LABELS.md reaches thirteen boards through the vendored mirror. ⚠️ And this is a discussion, so no needs-ruling label anchors a ladder to it — silence here has no timer at all, unlike a flagged issue.

---

Why the question moved

LABELS.md's paragraph is two rules, not one, and the exclusion sits at the end of the second:

An operator-owned issue normally reads ready + operator … ¶ An issue waiting on an external condition only the operator can resolve is operator-owned: it keeps ready and adds operator, and its quiet is legitimate under the same seven-day clock. This admits conditions the operator has the power to change … it does not admit a third party's unscheduled reply.

This has two available antecedents, and #573's own decisions point at the nearer one:

• D1 — "The rule as ruled: an issue waiting on an external condition that only the operator can resolve is operator-owed, and carries operator." That is the second sentence: an addition.
• D3 — "operator already covers the operator must do something. What the ruling adds is an external condition only the operator can resolve — … and not, for example, a third party's unscheduled reply."

D3 names the thing it bounds, and it is what the ruling adds. On that reading the exclusion is explicative, not restrictive — it spells out what the extension's own subject line ("only the operator can resolve") already excludes, which is exactly D3's stated job: "the sentence must distinguish the case it admits from the case it does not."

The honest counter-argument, which is why this is yours and not mine: D3's rationale sentence says "or the flag becomes the drawer everything unactionable goes into" — and the flag is the class, not the addition. D3 contains both framings. That is the whole ask.

What each answer does to the measured case

	A — bounds the addition	B — bounds the class
#20's label	already correct. It earns operator on TRIAGE.md's reach axis — no builder session reaches counsel — and never relied on D1's addition	no admissible label; the body's options 1–3 reopen as a follow-on decision
what ships	one scoping clause in LABELS.md naming which sentence the exclusion bounds	a widening, a taxonomy row, or an off-board home
what the criterion loses	the addition's legitimate-quiet allowance, so the seven-day nudge fires — and the body above already measured that limb as satisfiable by the operator	—

Three corrections to my own body, and one new control

1. ⚠️ "Both available shapes are wrong" is not established — it is a consequence of B, asserted in the body as a finding. Under A the shape #20 carries today is simply right, and there is no workaround in force to retire.
2. ⚙️ The measurement is not pin-local. The body cited the 0.7.8 blob. Re-run this pass in worktrees at both 0.7.8 and origin/main, owner_class_decision is byte-identical in behaviour: enhancement ready scope:template → ADD_BUILDER; + operator → KEEP; controls epic ready → KEEP, bare ready → ADD_BUILDER. So a ruling here binds every board, not one pin.
3. ⚙️ New control the body did not run: enhancement ready scope:template operator builder → REMOVE_BUILDER. The operator shape does not merely fail to add builder; it strips it. That strengthens the "less-wrong shape" argument under B and makes A's status-quo safer still.
4. 📎 The two surfaces that read the sentence as a class exclusion were one pass, not two independent readers — this thread's body and #20's criterion annotation were both written by triage on 2026-09-07. I am not claiming a census of misreadings; I am claiming one sentence with two antecedents.

The tripwire is unchanged and still dated

Under B, if anyone acts on the class reading and strips operator from #20, ADD_BUILDER writes builder onto a criterion that is nobody's build. That is inert only while heavy-duty/la-familia-incubator has no builder row — confirmed this pass, GET /labels/builder → 404 — and that row arrives with the same bootstrap=yes press #31's last criterion waits on. Under A nothing fires, because nothing moves.

One debt this comment does not pay

#20's criterion carries the superseded framing verbatim — "the less-wrong of two wrong shapes", "the workaround now in force". That text is triage's and is owed a correction on that board, pointing here; it is not corrected by this comment. It does already route the reader to this thread, so nobody following the citation is misled — but the sentence itself now overstates.

Verified this pass: #20 is open, unassigned, enhancement operator ready scope:template — read on its paginated label events (ready + operator set 2026-09-06T14:04:22Z, nothing since), not off its prose. Unassigned, so no attention is owed anywhere and setting one would be a board bug.

— triage, heavy-duty/ceremony, 2026-09-07.