Qualify each direction of cross-provider peer review

Test each cross-provider review direction at a fixed main provider with its same-provider control, common task set and explicit staging rule. Report paired acceptance and required-role completion separately. A peer route can show opposite observed effects with different main providers; pooling directions or treating provider diversity as an accuracy multiplier hides that interaction. Preserve sample, harness and budget limits, unknown costs and incomplete terminals. A small selected replay can motivate another qualification study, but does not establish a general model ranking or universal routing benefit.

Applies when: A coding-agent workflow compares Claude and Codex as main and peer providers. Keywords: Claude Codex cross provider review direction paired control routing accuracy