skills.rest
CarbeneAI
CarbeneAICOMMUNITY
Explore all skills by this author →
Website𝕏@CarbeneAIGitHub

codex

Review git diffs and plan files with Codex CLI for P1 findings.

9|1|Updated Apr 16, 2026
Search Tags:#security#cli#codex#telemetry#code-review#git-diff#adversarial-testing
One-click install
npx skills add https://github.com/CarbeneAI/Forge --skill codex-carbeneai
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: codex
Source: https://github.com/CarbeneAI/Forge/tree/main/.claude/skills/codex
Command: npx skills add https://github.com/CarbeneAI/Forge --skill codex-carbeneai

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Many repositories lack a fast, independent second-opinion that can both review changes and actively try to break them; humans miss adversarial edge cases and cross-model blind spots. Codex provides a distinct AI voice to run bounded code reviews, adversarial challenges, or consultative prompts that produce a clear pass/fail gate and actionable findings. This reduces uncertainty before shipping and surfaces issues other tools or reviewers may overlook.

Core Features & Use Cases

  • Three modes: review (diff-focused code review with a pass/fail gate), challenge (adversarial testing to find breakages), and consult (open-ended questions with session continuity).
  • Cross-model comparison: compare Codex findings with other reviews and report overlaps and unique issues.
  • Integration with git and plans: auto-detects diffs against a base branch, reviews plan files, and logs telemetry and review results for traceability.
  • Use Case: Run a codex review on your feature branch to detect P1 security or correctness findings and produce a gate verdict before merging.

Quick Start

Run the codex skill to review the current branch diff and produce a pass/fail gate with a detailed findings report.

Frequently Asked Questions about codex

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

FAQPage Schema
How do I run an independent adversarial code review on my git diff?▼

Adversarial code review targets git diffs against a base branch to actively find security, correctness, and design breakages. It produces a pass/fail gate based on P1 markers before merging.

Does AI code review work for finding edge cases humans miss?▼

AI code review provides a distinct second opinion that actively tries to break changes. It surfaces adversarial edge cases and cross-model blind spots that human reviewers often overlook during standard workflows.

What's the best way to compare code review findings across different AI models?▼

Cross-model comparison evaluates adversarial code review findings against other reviews. It reports overlapping issues and unique findings, helping reduce uncertainty before shipping by highlighting cross-model blind spots.

Do I need a specific CLI installed to run automated security code review?▼

Automated security code review requires the Codex CLI available on your system PATH. It enforces a filesystem boundary excluding ~/.claude and .claude/skills directories to maintain operational safety.

Can I use adversarial testing for open-ended codebase questions?▼

Adversarial testing supports an open-ended consult mode for ad-hoc prompts. This mode maintains session continuity while answering design and correctness questions without running a full diff-based review.

When should I not rely on an automated pass/fail gate for code review?▼

Automated pass/fail gates should not replace human review when evaluating complex design trade-offs. The gate strictly targets P1 security and correctness markers in git diffs and plan files, leaving architectural context to humans.

Related Skills

Explore More Software Engineering Skills→
Analog-LabsAnalog-Labs

codex

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

Official
Advanced
adolfoaranaes12adolfoaranaes12
2

quality-gate

Synthesize quality assessments into a go/no-go decision with YAML and Markdown reports.

Community
Advanced
jk278jk278

git-analyzer

Analyze git repositories to summarize commit history and file changes.

Community
Intermediate
BagelHoleBagelHole
46

sast-scanning

Performs static security testing on source code using Semgrep, CodeQL, and SonarQube.

Community
Intermediate