A Fixed-Direction Ephemeris Test of s^TX at the CMB Apex: Pre-Registered Prediction, Analysis Recipe, and a Provisional Bound from Public Data

Planetary-ephemeris constraints on the SME gravitational boost sector are degeneracy-limited: the published marginal on s^TX is ~8e-9 with near-unit correlations (Hees et al. 2015). A preferred-frame hypothesis (a de Sitter-Unruh modified-inertia reading of MOND with a0 = c^2 sqrt(Lambda/32pi) = 9.36e-11 m/s^2) FIXES the boost direction at the CMB dipole apex (RA,Dec = 167.94, -6.94 deg; component ratios locked), collapsing the test to ONE parameter with predicted amplitude |A| ~ 8.7e-10, negative sign. This note supplies: (1) the pre-registered prediction + kill condition (a dedicated 1-parameter fit null at sigma_A <= 4.3e-10 kills it; idealized archival floor sigma_A ~ 4e-11 -- decidable with data in hand, Mars-ranging dominated); (2) the concrete recipe for an ephemeris group (one added Bailey-Kostelecky boost acceleration, one fitted amplitude, fixed direction, no light-time changes); (3) a PROVISIONAL fixed-direction bound from public secular perihelion/node tables: A = (-2.5 +/- 5.9)e-10 (INPOP10a; (-0.9 +/- 6.4)e-10 with INPOP15a updates) -- the prediction sits at ~1.4 sigma with MATCHING SIGN, neither detected nor excluded, while direction-fixing alone sharpens the effective bound to |A| < 1.2e-9 (95%), ~2x better than published combined analyses, using only public tables; (4) a load-bearing scoping discovery: under the PER-BODY reading of the underlying relation (s ~ a0/2g per planet) the observable drifts fall 30-1000x below current sensitivity, so ephemerides test only the UNIVERSAL reading -- a fork not previously stated in the literature. CPT-even structure independently predicts exactly zero GW birefringence (kill-switch). Credit: formalism and methodology are Bailey-Kostelecky 2006 and Hees et al. 2015; the framework supplies only the direction-fixing hypothesis and target amplitude, is a one-parameter reframing of the MOND scale, and its earlier over-claims were publicly retracted (2026-06-23). Prediction and kill pre-committed to the public repository before any dedicated fit exists; the author accepts the kill as written. Reproducible: reviews/stx_target.py, reviews/stx_inpop_recipe.py (committed, exit 0).