Found by the triage seat, 2026-08-20 ~08:1xZ round, while acting on the half-state patrol view (#9857). Filed for the skills lane's own triage (.claude/ patrol tooling face). Not urgent — no state was damaged; the guidance was caught before being followed.
The live instance
The 07:54Z sweep reports:
H14#7917 — pm:blocking carried while NO open card's Blocked-by: line — body OR comment — targets it, judged against the full two-channel index — a stale derived cache. … Report-only: the remedy is the triage sweep's derivation pass dropping the label …
But #7917's dependent is objectstack-ai/objectui#4356 — open, pm:blocked, its Blocked-by: #7917 edge recorded on the objectui side (added 2026-08-19 when the #7917⇄ui#4356 cycle was converted to a chain). check-half-states.mjs sweeps this repo only, so its "full two-channel index" is complete over objectstack and structurally empty over the three sibling repos. The verdict is phrased as exhaustive ("judged against the full two-channel index") while the population it judged is not — and its prescribed remedy is precisely the action that would have broken the real edge: the triage derivation pass dropping pm:blocking from #7917, after which the selection order stops boosting the one card whose closure unblocks ui#4356.
This round kept the label and recorded the false positive here instead.
Why it is worth a card rather than a caveat
Directions (skills seat's call, not prescribed)
- Cheapest: H14's row text stops claiming completeness and names the boundary ("no dependent found in this repo; cross-repo dependents are not swept — verify before dropping"), making the triage-side cross-check explicit duty.
- Fuller: the patrol additionally greps the sibling repos' open blocked cards for
Blocked-by: lines targeting this repo (the reverse of what the unlock scan already does), at whatever cost ceiling the patrol tolerates. - Middle: H14 skips cards whose own thread records a cross-repo dependent (e.g. a
Blocking:/audit note left by the seat when the edge was created), turning the false positive into a documented suppression.
Refs: #9857 (the patrol view carrying the instance) · #8941 (the comment-channel gauge gap, same family) · #10102 (the backtick gauge gap, same family) · objectui#4356 (the live cross-repo dependent).
Found by the triage seat, 2026-08-20 ~08:1xZ round, while acting on the half-state patrol view (#9857). Filed for the skills lane's own triage (
.claude/patrol tooling face). Not urgent — no state was damaged; the guidance was caught before being followed.The live instance
The 07:54Z sweep reports:
But #7917's dependent is objectstack-ai/objectui#4356 — open,
pm:blocked, itsBlocked-by: #7917edge recorded on the objectui side (added 2026-08-19 when the #7917⇄ui#4356 cycle was converted to a chain).check-half-states.mjssweeps this repo only, so its "full two-channel index" is complete over objectstack and structurally empty over the three sibling repos. The verdict is phrased as exhaustive ("judged against the full two-channel index") while the population it judged is not — and its prescribed remedy is precisely the action that would have broken the real edge: the triage derivation pass droppingpm:blockingfrom #7917, after which the selection order stops boosting the one card whose closure unblocks ui#4356.This round kept the label and recorded the false positive here instead.
Why it is worth a card rather than a caveat
Blocked-by:lines in COMMENTS — blocked cards go invisible and sat 24h+ past discharged upstreams #8941 (body-only grep vs comment-channel lines) one level up: the reader trusts the index's completeness claim, and the incompleteness is invisible in the row itself.Blocked-by:edges are protocol-legal and common (contract-first splits mandate them:Blocked-by: <owner/repo>#<n>), so this is not an exotic case.Directions (skills seat's call, not prescribed)
Blocked-by:lines targeting this repo (the reverse of what the unlock scan already does), at whatever cost ceiling the patrol tolerates.Blocking:/audit note left by the seat when the edge was created), turning the false positive into a documented suppression.Refs: #9857 (the patrol view carrying the instance) · #8941 (the comment-channel gauge gap, same family) · #10102 (the backtick gauge gap, same family) · objectui#4356 (the live cross-repo dependent).