skills.rest
westkite1201
westkite1201COMMUNITY
Explore all skills by this author →
GitHub

codex

Wraps the OpenAI Codex CLI to review, challenge, or consult on code diffs.

2|Updated Mar 26, 2026
Search Tags:#diff analysis#code review#cli#codex#openai#adversarial testing#second opinion
One-click install
npx skills add https://github.com/westkite1201/garden-engine --skill codex-westkite1201
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: codex
Source: https://github.com/westkite1201/garden-engine/tree/main/.claude/skills/gstack/codex
Command: npx skills add https://github.com/westkite1201/garden-engine --skill codex-westkite1201

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires @openai/codex.

What problem does it solve? Getting a second opinion on code changes from the same AI that wrote them creates blind spots. This Skill invokes the OpenAI Codex CLI as an independent reviewer that analyzes your branch diff, adversarially probes for failure modes, or answers technical questions with session continuity. ## Core Features & Use Cases - Review Mode: Runs codex review against the base branch diff with a pass/fail gate based on critical [P1] findings, plus cross-model comparison with Claude's own review. - Challenge Mode: Adversarial analysis that hunts for edge cases, race conditions, security holes, and silent data corruption paths, with streamed reasoning traces. - Consult Mode: Ask Codex anything about the codebase with persistent session IDs for follow-up questions, including automatic plan-file review detection. - Use Case: Before merging a PR, run /codex review to get an independent gate verdict, then /codex challenge security to probe for exploit vectors the standard review missed. ## Quick Start Ask the assistant to run a codex review of the current branch diff against the base branch and report the pass/fail gate verdict.

Frequently Asked Questions about codex

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I get a second AI opinion on my code review?▼

Run the codex review mode, which executes `codex review --base <branch>` against your current diff. It returns a pass/fail gate based on critical [P1] findings and can compare results against Claude's own review for cross-model analysis.

How to adversarially test code for security holes with AI?▼

Use challenge mode with an optional focus like security. It prompts Codex to think like an attacker, hunting for injection vectors, auth bypasses, race conditions, and silent data corruption paths in your branch diff.

What happens if the Codex CLI is not installed?▼

The skill checks for the codex binary first and stops with an install message if missing. Install it with `npm install -g @openai/codex` or see the OpenAI Codex GitHub repository.

Can I continue a previous Codex conversation with follow-up questions?▼

Yes, consult mode stores a session ID in .context/codex-session-id. When a session exists, you are asked whether to continue the prior conversation with full context or start a fresh one.

How does the pass/fail gate in codex review work?▼

The gate parses the review output for severity markers. Any [P1] finding means FAIL, while output containing only [P2] markers or no findings means PASS. Results are persisted to a review log for plan-file reporting.

Related Skills

Explore More Software Engineering Skills→
eclipse-slmeclipse-slm
8

get-branch-changes

Determines merge-base and lists changed files for the current Git branch.

Official
Intermediate
bennybennisonbennybennison

review-issues

Review GitHub issues, pull requests, diffs, and bug reports for risks and missing context.

Community
Intermediate
no-sessionno-session
107

plan

Runs CEO, design, and engineering plan reviews sequentially with auto-decisions and a final approval gate.

Community
Advanced
eklyukineklyukin

codex

Obtains independent second opinions from OpenAI Codex CLI for planning, code review, and implementation review.

Community
Intermediate