codex

Review code diffs against the base branch using OpenAI Codex CLI modes.

Updated Mar 24, 2026

Search Tags:#cli #codex #session #ai-assistant #code-review #adversarial #dev-tools

Ready to run

npx skills add https://github.com/pistosmin/base-one --skill codex-pistosmin

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: codex
Source: https://github.com/pistosmin/base-one/tree/main/.claude/skills/gstack/codex
Command: npx skills add https://github.com/pistosmin/base-one --skill codex-pistosmin

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Codex CLI wrapper provides an independent, blunt second opinion on code changes by leveraging the OpenAI Codex CLI to evaluate diffs, prompts, and plans.

Core Features & Use Cases

Review mode to perform code reviews against the base branch, returning gate decisions and token estimates.
Challenge mode to adversarially probe changes for edge cases, security concerns, and robustness.
Consult mode to ask Codex anything about the project context, prompts, or plan questions, with session continuity.
Works across plan reviews and diff-driven workflows for teams needing neutral QC before shipping.

Quick Start

Run the codex skill to obtain a second opinion on the latest changes by selecting a mode (review, challenge, or consult) and supplying the diff or plan prompt.

Frequently Asked Questions about codex

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

FAQPage Schema

How do I get an adversarial code review on my latest git diff before merging?▼

An adversarial code review probes your latest git diff against the base branch to identify edge cases, security concerns, and robustness issues. Using challenge mode provides this adversarial analysis, returning blunt feedback on potential weaknesses before shipping your code changes.

What is a brutally honest second opinion on code changes and when do I need one?▼

A brutally honest second opinion is an independent, blunt evaluation of your code diffs that provides neutral quality control before shipping. You need one when your team requires an unbiased gate decision on code changes or project plans to catch issues internal reviewers might miss.

How do I use an AI assistant to review project plans and code diffs?▼

Review project plans and code diffs by selecting review mode to analyze changes against the base branch for gate decisions, or consult mode to query project context and prompts with session continuity for reproducible auditing and improvement.

Does the Codex CLI wrapper require session continuity for reproducible code reviews?▼

Yes, the Codex CLI wrapper ensures session continuity and telemetry for auditing and improvement. This reproducible session design allows you to track and verify code review decisions, challenge probes, and consultation queries consistently across multiple interactions.

What is the best way to challenge code changes for security and robustness concerns?▼

The best way to challenge code changes for security and robustness is using challenge mode, which adversarially probes your diffs against the base branch to identify edge cases and security concerns, returning blunt feedback on potential vulnerabilities before you ship.

Can I consult an AI assistant about project context and prompts with session continuity?▼

Yes, you can use consult mode to ask Codex anything about project context, prompts, or plan questions. Consult mode maintains session continuity, ensuring your queries are tracked for reproducibility and telemetry auditing across your development planning workflow.