How do I install the Codex Review skill?

Run npx skillmds@latest add mustafahasanain/codex-review in your terminal (requires Node.js), paste this page's agent-chat prompt into Claude, Cursor, or any MCP-connected agent, or download the SKILL.md file and copy it into your agent's skills directory.

Is Codex Review safe to use?

This skill has not completed SkillMD's automated safety review yet. SkillMD never runs a skill's scripts for you; review the SKILL.md before installing.

Which AI agents work with Codex Review?

This skill is tagged as working with Claude Code, Claude.ai, OpenAI Codex. SKILL.md is an open format, so most agents that read a skills directory can load it too.

Is Codex Review free to use?

Yes. Installing skills from SkillMD is free, and the skill stays under its author's original license.

Who published Codex Review?

mustafahasanain (@mustafahasanain) published this skill. Their other Agent Skills are listed on their SkillMD profile.

Codex Review by mustafahasanain — Agent Skill

Name: Codex Review
Author: mustafahasanain

Codex Review

Codex review lifecycle - build the review bundle, generate the prompt, parse APPROVED/CHANGES_REQUIRED results, run the review-fix loop, and invalidate approval via the implementation fingerprint.

mustafahasanain fa3bdd4 5 files · 12.2 KB Updated Sep 17, 2026

File contents

Codex Review Skill

Manages the external Codex review of a completed implementation. Codex reviews only — it never modifies files; the bundle says so explicitly.

When to Use

review action (stage implementation/review_fixes, completion gate passed) → bundle + prompt (references/review-prompt-generation.md).
apply-review action (stage codex_review) → parse result (references/review-result-parsing.md).
Stage review_fixes → fix loop (references/review-fix-loop.md).
Any post-approval change → (references/approval-invalidation.md).

Inputs

Passing completion gate; docs/ai-context/TASK_PLAN.md; the implementation summary at docs/ai-context/implementation-summary.md; Git state; for apply-review, the raw Codex result text or file.

Outputs

docs/ai-context/reviews/round-NNN-prompt.md (via bin/frappe-workflow review bundle, which secret-scans and embeds the fingerprint).
docs/ai-context/reviews/round-NNN-result.md — the original result, preserved verbatim.
State updates: round, prompt/result paths, fingerprint, status (pending → changes_required → … → approved), approved_at.

Preconditions

Bundle: validate completion-gate passes.
Result parsing: a prompt for the current round exists and state is codex_review.
Approval acceptance: current fingerprint == recorded fingerprint.

Stopping Conditions

APPROVED accepted → finalization (feature-changelog + project-context updates, review record) → validate finalization-gate → ready_for_commit. Done.
CHANGES_REQUIRED → review_fixes; loop continues until approved.
Malformed result → reject with the format requirements; stay in codex_review; never guess a status.

Prohibited

Generating a prompt without a passing completion gate.
Accepting any status other than exactly APPROVED / CHANGES_REQUIRED.
Inventing, dropping, or "interpreting away" findings; the original result file is never edited.
Accepting approval whose fingerprint mismatches the working tree.
Applying a finding without validating it against the repository first.

Shared Rules

shared-workflow-rules.md, security-rules.md.

---name: codex-reviewdescription: Codex review lifecycle - build the review bundle, generate the prompt, parse APPROVED/CHANGES_REQUIRED results, run the review-fix loop, and invalidate approval via the implementation fingerprint.---# Codex Review SkillManages the external Codex review of a completed implementation. Codexreviews only — it never modifies files; the bundle says so explicitly.## When to Use- `review` action (stage `implementation`/`review_fixes`, completion gate passed) → bundle + prompt ([references/review-prompt-generation.md](references/review-prompt-generation.md)).- `apply-review` action (stage `codex_review`) → parse result ([references/review-result-parsing.md](references/review-result-parsing.md)).- Stage `review_fixes` → fix loop ([references/review-fix-loop.md](references/review-fix-loop.md)).- Any post-approval change → ([references/approval-invalidation.md](references/approval-invalidation.md)).## Inputs- Passing completion gate; `docs/ai-context/TASK_PLAN.md`; the implementation summary at `docs/ai-context/implementation-summary.md`; Git state; for `apply-review`, the raw Codex result text or file.## Outputs- `docs/ai-context/reviews/round-NNN-prompt.md` (via `bin/frappe-workflow review bundle`, which secret-scans and embeds the fingerprint).- `docs/ai-context/reviews/round-NNN-result.md` — the original result, preserved verbatim.- State updates: round, prompt/result paths, fingerprint, status (pending → changes_required → … → approved), approved_at.## Preconditions- Bundle: `validate completion-gate` passes.- Result parsing: a prompt for the current round exists and state is `codex_review`.- Approval acceptance: current fingerprint == recorded fingerprint.## Stopping Conditions- `APPROVED` accepted → finalization (feature-changelog + project-context updates, review record) → `validate finalization-gate` → `ready_for_commit`. Done.- `CHANGES_REQUIRED` → `review_fixes`; loop continues until approved.- Malformed result → reject with the format requirements; stay in `codex_review`; never guess a status.## Prohibited- Generating a prompt without a passing completion gate.- Accepting any status other than exactly `APPROVED` / `CHANGES_REQUIRED`.- Inventing, dropping, or "interpreting away" findings; the original result file is never edited.- Accepting approval whose fingerprint mismatches the working tree.- Applying a finding without validating it against the repository first.## Shared Rules[shared-workflow-rules.md](../../references/shared-workflow-rules.md),[security-rules.md](../../references/security-rules.md).

Frequently asked questions