skills.rest
centminmod
centminmodCOMMUNITY
Explore all skills by this author →
WebsiteGitHub

consult-codex

Orchestrates parallel code analysis between OpenAI Codex CLI and Claude code-searcher with citation comparison.

2.6k|246|Updated Jul 8, 2025
Search Tags:#citations#Debugging#dual-ai#second opinion#Codex CLI#Code review#Code Analysis
One-click install
npx skills add https://github.com/centminmod/my-claude-code-setup --skill consult-codex
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: consult-codex
Source: https://github.com/centminmod/my-claude-code-setup/tree/main/.claude/skills/consult-codex
Command: npx skills add https://github.com/centminmod/my-claude-code-setup --skill consult-codex

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires jq, codex.

What problem does it solve?

Getting a second opinion on complex code questions normally means manually querying multiple AI tools and reconciling their answers yourself. This Skill automates dual-AI consultation by dispatching the same structured prompt to OpenAI Codex and Claude's code-searcher agent in parallel, then synthesizing their citation-backed findings into one comparison report.

Core Features & Use Cases

  • Parallel Dual-AI Dispatch: Sends an enhanced prompt with line-number citation requirements to both Codex CLI and the code-searcher subagent simultaneously, cutting total response time.
  • Structured Comparison Report: Produces raw responses from both agents plus a comparison table, agreement level, corroboration-bucketed findings, and a synthesized recommendation.
  • Resilient Execution: Handles broken nvm-installed Codex binaries, missing jq, hung CLI processes via timeout watchdogs, and degraded single-agent runs when one agent fails.
  • Use Case: When debugging a difficult issue or reviewing architecture, ask a code question and receive two independent analyses with file:line citations, bucketed as corroborated, solo, or disputed findings.

Quick Start

Ask a code question like "consult codex on how the authentication flow works in this repository" to get a dual-AI citation-verified analysis.

Frequently Asked Questions about consult-codex

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

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

Invoke this Skill with your code question and it dispatches the same enhanced prompt to OpenAI Codex and Claude's code-searcher agent in parallel. You receive both raw responses plus a synthesized comparison with file:line citations and an agreement assessment.

What is the difference between Codex and code-searcher for code analysis?▼

Codex runs as an external CLI in read-only sandbox mode using its configured model, while code-searcher is a Claude subagent using Read, Grep, Glob, and Bash directly. The Skill compares their findings side by side, highlighting corroborated, solo, and disputed results.

What dependencies are required to run dual-AI code consultation?▼

jq is a hard dependency for parsing Codex JSONL output, and the codex CLI must be resolvable via PATH or an nvm node installation. A timeout or gtimeout binary from coreutils is recommended to bound hung Codex dispatches.

What happens if Codex fails or times out during a consult run?▼

The Skill detects failures through empty output guards and timeout exit codes, then presents the successful agent's response as a degraded single-AI run. The failure is noted in the report and the comparison table is omitted since there is nothing to compare.

When should I not use dual-AI code consultation?▼

Skip it for simple syntax questions, basic file lookups, or straightforward documentation queries where a single AI suffices. The Skill is designed for complex analysis, debugging, architecture questions, and code reviews where two perspectives add value.

Related Skills

Explore More Software Engineering Skills→
hiddink-aihiddink-ai
1

codex-exec

Execute OpenAI Codex CLI prompts in ephemeral environments with JSON output.

Official
Advanced
allcllallcll
26

allcll-analysis

Analyze backend code and system behavior to identify root causes and risks.

Official
Intermediate
shakhovskiya-createshakhovskiya-create

tooling-python

Run Ruff, MyPy, Bandit, Pip-audit, Radon, Vulture, and Mutmut audits on Python projects.

Community
Advanced
austintheriotaustintheriot
1

consult

Route user questions to a single consult-capable subagent for direct expert advice.

Community
Basic