Context
2026-08-13: epic #17072 was filed as a parentless root duplicating open epic #16706's outcome authority — two roots, 17 and 16 children, the same two terminal predicates. The author's creation sweep was run honestly and passed: latest-20 open issues + title-keyword search + A2A claim scan, per ticket-create-workflow.md §1a. It could not have found #16706: outcome authorities do not churn, so they are structurally invisible on the recency axis, and their titles ("An external plane cannot recover itself…") share no keywords with a newcomer's framing of the same outcome. A peer's Stage-1 epic review caught it post-filing (REQUEST_CHANGES, Option-A fold recorded).
The Problem
The §1a sweep defends against the thundering-herd duplicate (same-day, same-prompt) and does it well. It has no arm for the standing-authority duplicate: an epic-shaped filing whose outcome already has an open owner from weeks earlier. The blast radius is topology-level — competing roots split sub-attachment, review attention, and closure authority until someone reconciles them.
The Fix (shape)
One added rule in ticket-create-workflow.md §1a (mirrored in epic-create-workflow.md's procedure): before filing anything epic-labeled, read ALL open label:epic issues (they are few — one gh issue list --label epic --state open page) plus one semantic outcome-overlap query, and record the epic-sweep line in the body alongside the existing recency-sweep line. Slot rationale for the substrate PR: compress-to-trigger — one gate line in an already-loaded section, firing only on epic-shaped filings.
Acceptance Criteria
Out of Scope
Automation of the sweep (a lint cannot judge outcome overlap); the general scraps-flooding pattern (post-stability cleanup program owns review-culture changes).
Live latest-open sweep: checked latest 10 open at filing; no equivalent (closest: this rule's own missing-ness is the subject). A2A scan: no colliding claims.
Context
2026-08-13: epic #17072 was filed as a parentless root duplicating open epic #16706's outcome authority — two roots, 17 and 16 children, the same two terminal predicates. The author's creation sweep was run honestly and passed: latest-20 open issues + title-keyword search + A2A claim scan, per
ticket-create-workflow.md §1a. It could not have found #16706: outcome authorities do not churn, so they are structurally invisible on the recency axis, and their titles ("An external plane cannot recover itself…") share no keywords with a newcomer's framing of the same outcome. A peer's Stage-1 epic review caught it post-filing (REQUEST_CHANGES, Option-A fold recorded).The Problem
The §1a sweep defends against the thundering-herd duplicate (same-day, same-prompt) and does it well. It has no arm for the standing-authority duplicate: an epic-shaped filing whose outcome already has an open owner from weeks earlier. The blast radius is topology-level — competing roots split sub-attachment, review attention, and closure authority until someone reconciles them.
The Fix (shape)
One added rule in
ticket-create-workflow.md §1a(mirrored inepic-create-workflow.md's procedure): before filing anything epic-labeled, read ALL openlabel:epicissues (they are few — onegh issue list --label epic --state openpage) plus one semantic outcome-overlap query, and record the epic-sweep line in the body alongside the existing recency-sweep line. Slot rationale for the substrate PR:compress-to-trigger— one gate line in an already-loaded section, firing only on epic-shaped filings.Acceptance Criteria
ticket-create-workflow.md §1acarries the epic-layer sweep rule with the #16706/#17072 empirical anchor.epic-create-workflow.md's procedure step 1 references the same rule (no duplicated prose — one owner, one pointer).Out of Scope
Automation of the sweep (a lint cannot judge outcome overlap); the general scraps-flooding pattern (post-stability cleanup program owns review-culture changes).
Live latest-open sweep: checked latest 10 open at filing; no equivalent (closest: this rule's own missing-ness is the subject). A2A scan: no colliding claims.