← ClaudeAtlas

codex-reviewlisted

A standalone adversarial PLAN-review loop where Claude Code (builder) and OpenAI Codex (read-only critic) tag-team an implementation plan before any code is written. Use this when you ALREADY have a plan or a clear idea and just want the cross-model stress-test — no requirements interview first. Claude drafts/loads the plan into PLAN.md, Codex reviews it in a read-only sandbox and returns VERDICT:APPROVED or VERDICT:REVISE, Claude revises and re-submits to the SAME Codex session (context preserved) until APPROVED or a configurable MAX_ROUNDS cap is hit. Human approves the converged plan before code. Use when the user says "/codex-review", "codex review my plan", "have Codex review my plan", "argue this plan with Codex", "adversarial plan review", "make Claude and Codex argue/fight over the plan", or is about to build something high-stakes (auth, schema, concurrency, migrations, payments) and wants a second-model sanity check on the PLAN before implementation. For a guided requirements interview BEFORE the rev
will-pagane/claude-setup · ★ 4 · Code & Development · score 72
Install: claude install-skill will-pagane/claude-setup
# Codex-Review — Adversarial Plan-Review Loop Two models, one plan, a bounded argument. **Claude is the builder and orchestrator. Codex is a read-only critic** that can read the repo and the plan but cannot touch a single file. They communicate strictly through this run's plan file (`$PLAN_FILE`, under a timestamped `$RUN_DIR`) + a Codex session that persists across rounds. The human enters at exactly two points: kickoff and final sign-off. This is a **deliberate, high-stakes tool** — reach for it on auth, data models, concurrency, migrations, payments, anything expensive to get wrong. Skip it for obvious/cheap work. ## Prerequisites (verify once, fast) - Codex CLI installed and recent: `codex --version` (need ≥ 0.130; the default `gpt-5.5` model errors on older CLIs). - Codex authenticated: a prior `codex login` (ChatGPT account is fine). If a run returns an auth/model error, surface it to the user — do not silently retry. - Do NOT pin `-m` unless the user asks. The user's `~/.codex/config.toml` default model is used. Pinning `gpt-5.x-codex` variants fails on ChatGPT-account auth. - **Sandbox flag differs between the two commands.** `codex exec` accepts `-s read-only`. `codex exec resume` does NOT — it rejects `-s` ("unexpected argument"). On resume you MUST force read-only via `-c sandbox_mode="read-only"`, because `config.toml` may default `sandbox_mode` to `danger-full-access` (+ `approval_policy="never"`) — which would let Codex WRITE files mid-loop. This is the sing