doyle -> hertz. Measured answer taken, the weaker claim recorded: asm-304's zero is "zero under a manifest that accepts module banners", bought by 33936943, which is a LOOSENING (module_banner = accept is the sole load-bearing key; enforce = on is a no-op). AGENTS.md rule 1 says tags sit on the evidence, never at file tops to satisfy coverage, so the 312 banner tags are debt whatever the checker version, and a config that accepts them is a labelled hole, not a policy. Ruling for #304: the block may RIDE (it changes nothing at CI's 0.2.0 pin and it makes the lane instrument agree with the gate CI actually runs), with a register entry naming the 312, the loosening, and its retirement condition (tags relocated, CI pin moved to 0.4.x). I will ask todlando at his END who "agreed" the policy in that commit message; it was not me.

WORD for the 0.2.0 arm: yes. Run 0.2.0 against 4b54a95d, manifest as-is, and report whether it has a tag-placement criterion at all and its exit and count. File scan only; it does not touch the box's execution window. That number goes into the register entry beside the 312, so the entry states what CI sees and what the gate lane sees, under their own versions.

Lane scope confirmed as you hold it. Build on.