codex-oracle

Name: codex-oracle
Availability: InStock
Author: Recall

Run Codex CLI as an independent agent to review codebase diffs.

1|Updated Apr 28, 2026

Search Tags:#cli #verification #codex #ai-assistant #code-review #diff-analysis #independent-agent

Ready to run

npx skills add https://github.com/recallnet/polymarket-cross-sectional-momentum --skill codex-oracle-recallnet

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: codex-oracle
Source: https://github.com/recallnet/polymarket-cross-sectional-momentum/tree/main/.agents/skills/codex-oracle
Command: npx skills add https://github.com/recallnet/polymarket-cross-sectional-momentum --skill codex-oracle-recallnet

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires curl, jq, and includes scripts (resource) components.

What problem does it solve?

Automates independent code review and verification by running Codex as a separate agent that can read the codebase, review diffs, and produce precise analyses.

Core Features & Use Cases

Independent code review with Codex using the codex CLI to inspect the codebase and diff changes
Supports blind and directed modes for flexible review and cross-model verification
API fallback via scripts/call-openai.sh for pure reasoning tasks when file access is not required
Integrates with multi-model workflows and adversarial checks to ensure robust verification

Quick Start

Install the codex CLI, authenticate, and run codex review on your uncommitted changes to start an independent code review.

Frequently Asked Questions about codex-oracle

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

FAQPage Schema

How do I automate independent code review on my uncommitted diffs?▼

Automate independent code review by running Codex as a separate agent to read the codebase and review diff changes. It supports CLI-based workflows to inspect uncommitted changes and produce precise analyses.

What is the difference between blind and directed modes for AI-assisted code review?▼

Blind and directed modes in AI-assisted code review provide flexible review strategies and cross-model verification. Blind mode reviews without prior context, while directed mode targets specific diff changes for precise analysis.

Can I use Codex CLI for cross-model verification and adversarial checks?▼

Yes, Codex CLI supports cross-model verification and adversarial checks through the /verify pathway. It runs as an independent agent to enable gate-level checks across multiple models for robust verification.

Do I need curl and jq installed to run Codex Oracle for diff analysis?▼

Yes, curl and jq are required dependencies to run Codex Oracle for diff analysis. Curl handles API fallback requests via scripts, while jq processes JSON responses for code review workflows.

When should I use the API fallback script instead of CLI-based code review?▼

Use the API fallback script via scripts/call-openai.sh for pure reasoning tasks when file access is not required. CLI-based code review is preferred when the agent needs to read the codebase and inspect diff changes.

What is the best way to set up independent agent code review for multi-model workflows?▼

The best way to set up independent agent code review for multi-model workflows is running Codex as a separate agent via CLI. It supports cross-model integration and adversarial checks through the /verify pathway for gate-level verification.