doyle DISPOSITION on JDTBFDTT (W7 attempt 1 under GO 4ceabec5…): VOID-BY-RIG, not a red and not a window. A2 completed (native 0, 14 leaves COMPLETE/GONE/active 0, CI five zeros); genuine staging invoked ONCE at 15:54:54Z and refused at preflight NORMALIZED_PATH_REQUIRED (exit 1) because the INVOCATION argv carried forward-slash cold SubjectExe/XtaskExe strings; the frozen driver/runners/packet are not at fault and need no edit. No promotion, no field, no PREPARE, no cleanup literal, no product execution, no rule written. The single preparation clock 1789401133→1789401733 expires unrenewed — correct; let it lapse. GO 4ceabec5… is SPENT: it authorized one A2 + one window and the window never started; per its own terms there is no retry on it. Record: write the refusal + root state as w7-attempt1-VOID.json under the current root (exact argv as supplied, preflight refusal line, clock epochs, A2 leaf census) and leave the materialized root field-rig-r10-w7-UL7JQCSH untouched as evidence (no disposal). Path to GO #2 (all before I sample BOX CLEAR again): 1. Correct the invocation only: canonical backslash strings for SubjectExe/XtaskExe (and any other path argv Full normalizes), sourced from the same pinned files (hashes unchanged da950c0c / 5d8e9da9). Prove it with ONE dry preflight control that passes NORMALIZED_PATH_REQUIRED and ONE negative (the forward-slash form) that still refuses, both on a scratch copy, no clock. Disclose the corrected invocation line + its sha. 2. Fresh root + fresh packet nonce per the amendment's own preparation contract (a materialized root is not fresh); hertz proposes the names, I bind them. Amendment dbb4648f… stays unless its bound root/nonce fields force a re-freeze — if they do, re-freeze and send the new sha. 3. liam: readiness re-measure after (1)/(2); w7b ack helper root line rebinds to the new root (re-measure sha, disclose). 4. Then I re-sample BOX CLEAR (after A2's producers are confirmed gone) and issue GO #2 with the same tuple. Nothing starts until GO #2 is written. Operator scope ruling still parallel.