CAMS Protocol R (CAMS-R-2026-001): preregistration, single-run results, and rubric erratum final

Protocol R2 — Holdout Custodian Re-declaration Date: 2026-09-20 Custodian: Kimi Work agent, local instance (Kari McKern, ComplexityWorkz) Study: CAMS-R2-2026-001 (successor to CAMS-R-2026-001) Declaration I accept custodianship of the protocol-R2 holdout vintage under the same terms as the R1 custodianship, re-stated for this vintage: The R2 holdout roster is thirteen societies: India, Chile, Greece, Netherlands, New_Zealand, Singapore, Ukraine, Laos, Burma, DRC, Sweden, Latvia, and United_Kingdom — the last admitted as the registered conditional thirteenth after its re-scored panel passed the §8 constant-series scan at sealing (mechanical rule, no discretion exercised). The vintage is sealed in holdout_custody_manifest_R2.json (sha256 8476b653e6a7651092fdad603a4881b264a7def51e18e681b96ec53baffb286d). New_Zealand, Sweden, Latvia and United_Kingdom panels contain re-scored erratum-E-2026-001 series only; all other node-metrics are byte-identical to the R1 panels. Provenance and the blindness limitation are disclosed in rescore_R2/PROVENANCE.md. The holdout side of r2_analysis.py will be executed exactly once, with its output logged. No exploratory holdout computation, no peeking, no re-runs. The only pre-run computation on the vintage was the registered sealing constant-series scan (no correlations were computed). Development vectors are frozen in r2_dev_vectors_frozen.json; the raw vector reproduces the R1 frozen vector exactly (assert-verified). Verdict language is pre-registered (STRONG_PASS / SYSTEMIC_PASS / WEAKENED / FAIL) and will be reported exactly as it falls. Hashes at sealing Item sha256 (first 16) R2 rubric bundle 68b2671ee4eedea8 r2_analysis.py 6a12857900c54c60 R2 roster manifest (25 entries) 0cdb5a9b25a4bf21 → superseded by sealed manifest below R2 holdout custody manifest 8476b653e6a76510 Signed: Kimi Work agent instance, 2026-09-20