grill-me-codex

Two AI models harden your plan — then swap jobs to build it. A family of Claude Code skills that close the two gaps in AI-assisted coding: the gap between you and Claude (do we agree on what to build?) and the gap between Claude and the quality of what it produces (is the plan actually correct — and how would you even know?).

Act 1 grills you to lock the plan. Act 2 hands that plan to OpenAI Codex — a rival, cross-provider model — which adversarially tears it apart over several rounds until both models sign off. Act 3 (optional) flips the roles: Codex writes the code from the frozen plan while Claude reviews the diff like a contributor PR. Cross-model checks in both directions — nobody grades their own work.

Why a second model? Because the same model that plans the build and writes the build can't be trusted to grade its own work — it's an echo chamber. A different provider catches what Claude structurally can't see in itself.

Built on Matt Pocock's grill-me / grill-with-docs skills (MIT) — Act 1 is his work; the iterative cross-model Codex review (Act 2) and the role-flipped build (Act 3) are the additions. Act 3's delegation pattern is adapted from Peter Steinberger's codex-first.

The skills

Skill	Act 1	Act 2	Act 3	Use when
grill-me-codex	Claude interrogates you one question at a time until the decision tree is resolved	Codex adversarial review loop	optional → codex-build	Planning from scratch + want alignment AND a second-model check
grill-with-docs-codex	Same, but challenges your plan against your project's CONTEXT.md glossary + writes ADRs inline	Codex review loop	optional → codex-build	Same, in a project with established terminology / architecture decisions
codex-review	— (you already have a plan)	Codex review loop	optional → codex-build	You have a plan and just want the cross-model stress-test
codex-build	—	—	Codex implements the frozen plan; Claude verifies	You have a reviewed spec and want the second model to type it

How Act 2 works (the review)

1. Claude writes the locked plan to PLAN.md and starts a log at PLAN-REVIEW-LOG.md.
2. Round 1: Codex reviews the plan in a read-only sandbox and returns VERDICT: APPROVED or VERDICT: REVISE.
3. Rounds 2..N: Claude revises; the same Codex session is resumed so it remembers its prior critiques and only checks whether they're addressed.
4. Bounded by MAX_ROUNDS (default 5). Terminates on APPROVED or the cap (a flagged deadlock beats a fake "approved").
5. You gate twice only: kickoff, and final sign-off before any code. Codex is read-only every round and never writes a file.

Two artifacts: PLAN.md (the clean final plan — the what) and PLAN-REVIEW-LOG.md (the full round-by-round argument — the why).

How Act 3 works (the build — roles flip)

1. After you sign off the plan, codex-build hands PLAN.md to Codex as a frozen spec — Codex gets full write access (--yolo) and implements it end-to-end, running tests as it goes. Requires a clean git tree first, so its diff is isolatable and revertible.
2. Claude — now the critic — reads the full diff like a contributor PR and runs the proof test itself. Codex's claims are advisory; Claude's own run is the proof.
3. Problems go back to the same Codex session as precise fix rounds, capped at MAX_FIX_ROUNDS (default 2) — after that Claude takes over and finishes by hand rather than ping-ponging.
4. You gate once more: the diff sign-off. Claude writes the commit; Codex never commits.
5. Build rounds append to the same PLAN-REVIEW-LOG.md, so one artifact tells the whole story: grilled → reviewed → built → verified.

Bonus: Codex sessions have a native image-generation tool (ChatGPT-account backed, no API key). A spec can include "generate these image assets yourself" steps — sprites, textures, backdrops — with exact paths and dimensions in the build contract.

Install

Copy the skill folders into your Claude Code skills directory:

# macOS / Linux
cp -r skills/* ~/.claude/skills/

# Windows (PowerShell)
Copy-Item -Recurse skills\* $env:USERPROFILE\.claude\skills\

Then invoke in Claude Code: /grill-me-codex, /grill-with-docs-codex, /codex-review, or /codex-build.

Prerequisites

• Codex CLI ≥ 0.130 — npm install -g @openai/codex@latest (older versions error on the default gpt-5.5 model).
• Authenticated Codex — run codex login once (a ChatGPT account works; Free/Plus/Pro/Max all fine).
• Don't pin a model — ChatGPT-account auth rejects gpt-5.x-codex model variants; the skills use your config default.

Tunables

Skill	Var	Default	Meaning
review skills	MAX_ROUNDS	5	Hard cap on review rounds
review skills	PLAN_FILE	PLAN.md	Where the plan lives
all	LOG_FILE	PLAN-REVIEW-LOG.md	The argument transcript
codex-build	SPEC_FILE	PLAN.md	The frozen spec Codex implements
codex-build	MAX_FIX_ROUNDS	2	Fix rounds before Claude takes over
codex-build	PROOF_CMD	from spec	Exact test command that counts as proof

Pass e.g. rounds=3 when invoking to override.

Safety

Review skills (Acts 1–2): Codex runs read-only every round — -s read-only on the first call, -c sandbox_mode="read-only" on every resume (the resume subcommand doesn't accept -s, and without forcing read-only it would inherit your config.toml sandbox default, which may be danger-full-access). The skills handle this for you. No code is ever written until you approve the final plan.

codex-build (Act 3) deliberately inverts this: Codex gets full write access — which is exactly why the skill gates it hard. Clean git tree required before launch (diff isolation + clean revert), Claude reads every line of the diff and runs the proof itself, fix rounds are bounded, commits are human-gated and Claude-authored. Resume calls need the long flag --dangerously-bypass-approvals-and-sandbox (resume has no --yolo) — and always resume by explicit thread_id, never --last: a wrong or missing id can silently land in a different session.

Credits

• Act 1 (grill-me, grill-with-docs) © Matt Pocock — https://github.com/mattpocock/skills (MIT). See each skill's THIRD-PARTY-NOTICES.md.
• Act 3's Codex-as-builder pattern adapted from Peter Steinberger's codex-first.
• Act 2 (iterative Codex review), Act 3 (codex-build), and packaging by Chase AI.
• Want to go deeper? The Claude Code Masterclass and a community of builders shipping with agentic AI live inside Chase AI+.

License

MIT — see LICENSE.