Budget-aware twin of plan-opus-execute-codex — Opus plans, Sonnet 4.6 explores AND builds, GPT-5.5 reviews cross-family. Use when the scarce GPT-5.5/Codex budget must be conserved for review only, when you want the token-heavy build on the Claude/Sonnet quota, or when `omp usage` shows the Openai Codex account near its cap — i.e. "execute with sonnet", "build with claude, review with codex", "conserve codex budget". Not for when Codex budget is healthy and you want GPT-5.5 to build (use plan-opus-execute-codex), or when you want GPT-5.5 to design the plan (use plan-codex-execute-opus).
