Twinned from https://github.com/cfsh/ninefold/issues/964 by tools/beads/import_github.py's reconcile pass.
design_docs/009-the-games-audit-lane.md (PR #963) — accepted by Aedan in chat, 2026-08-15 (*"for now, I agree lets just unlock changes within a single game from needing approval"*), chosen over PR #934's draft (#935 closes with it). Scope grew on his ask (same chat) to include the label-vocabulary cleanup (#967 closed as absorbed).
Recommendation: narrow the Tier-1 perimeter's R1 rule itself — a PR whose whole diff sits inside one *existing* games/<name>/ merges as ordinary Tier 1 (reviewer converged at head, standard gates), no new authority, no audit tooling. Plus the taxonomy: tier:* → risk:{routine,infra,feel} (word *tier* reserved for hacky-vs-hardened); a new informational pre-merge approval:auto/approval:human label projecting classify() onto every open PR; merged:* and --audit scrapped altogether — the required merge-authority check server-enforces what their label scan detected after the fact.
Shape: four independent one-PR tasks — T1 the single-game exception in perimeter.classify() (the governed activation PR; Aedan's approval of it is the trust decision, reverting it the kill switch), T2 actionable refusal reasons, T3 the risk rename, T4 approval:* in / merged:* + --audit out. Full task cards in the doc's §8.
Not claimed. Not labelled.
None.
No comments.