CDN-247 · The First Score — GCFT Dark-Sector Book Arc (Art Edition)

CDN-247 executes the locked CDN-242 seven-term κ-proxy (b₆ = +21. 98 at ln λ₀ = ln 17. 5, bᵣefined of record from cdn242ᵣesidualᵣeductionᵣesults. json) against the CDN-246 v2 holdout sample (15 clusters, audit verdict PROCEED). The CDN-242 population-averaged non-GCFT bound HOLDS at 0. 215% (target ≤ 0. 29%, PASS-band edge 0. 39%): zero single-cluster cap violations, mean dark-sector closure 98. 97%. Per-cluster landings against the CDN-243 prediction bands: 10 of 15 clusters in band (66. 67%), five below their assigned class's lower band edge. The shortfall falls between the CDN-243 PASS threshold (≥80% in band) and the FAIL threshold (<60%), triggering VERDICT: CONFLICT. The bound is preserved; the prediction-band specification is diagnostically incomplete on this sample. Five out-of-band clusters all land BELOW their lower band edge: Abell 267 (S, R/ξ=0. 025), Abell 586 (S, 0. 061), Abell 1413 (S, 0. 042), MS 0451. 6-0305 (S, 0. 006), Abell 2219 (P, 0. 021). The proxy over-closes on the relaxed-leaning v2 mix relative to the CDN-243 S-band calibration, which was tuned on the CDN-242 sample's mass / morphology distribution. None of these is a refutation: every cluster lands at or below its predicted band, none above; the population bound holds; no cluster exceeds its single-cluster cap. Class composition: 10 E, 4 S, 1 P, 0 F. The absence of F-class clusters means the Class-F super-ceiling claim is unaddressed (neither confirmed nor refuted) by this draw. The CDN-242 dark-sector residual decomposition survives intact as a working framework; the band specification needs calibration diagnosis on the relaxed end. Cluster characteristic-parameter vectors (M₂00, fgas, mergerₒffset, stellargasᵣatio, turbₚroxy, morphₚroxy, relaxed flag, lensingₒffset, z) are modelled-from-literature values in the same convention as the CDN-240 CLUSTERS dict; per-cluster ref field cites the primary literature source for the cluster's type assignment. CDN-248 must replace these vectors with primary-literature pulls before promoting any per-cluster claim. CDN-248 must NOT refit the κ-proxy in response to CONFLICT — that would erase the holdout's value as an independent test. Diagnostic priorities: (a) histogram CDN-242 S residuals against v2 S residuals to determine whether the S-band lower edge is a CDN-242-sample calibration artefact; (b) consider a targeted F-class draw to address the unaddressed super-ceiling regime; (c) replace modelled cluster vectors with primary-literature pulls. Art Edition with star-field design system (seed 247). Lock: CDN247HOLDOUTEXECUTIONLOCKED,