codex-help

Generate structured Codex diagnoses with actionable next steps for blocked Claude tasks.

Updated Mar 26, 2026
One-click install
npx skills add https://github.com/mrkhachaturov/rkstack --skill codex-help
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: codex-help
Source: https://github.com/mrkhachaturov/rkstack/tree/main/skills/codex-help
Command: npx skills add https://github.com/mrkhachaturov/rkstack --skill codex-help

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Claude often encounters blockers during tasks and needs expert, structured guidance to unblock work. This skill provides a controlled, iterative Codex-help loop that yields a diagnosis, concrete next steps, verification, and loop management, while Claude remains in command.

Core Features & Use Cases

  • User-invocable, read-only Codex assistance that returns a structured analysis, a clear diagnosis, and actionable steps.
  • Supports multi-round iteration with a persistent Codex thread to converge on a resolution and verify outcomes.
  • Includes safety constraints that keep Claude in control while Codex provides advisory guidance.

Quick Start

Invoke /codex-help to start Codex-assisted debugging rounds and iteratively unblock the task.

Frequently Asked Questions about codex-help

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

FAQPage Schema
How do I get structured debugging guidance when Claude hits a blocker?

Structured debugging guidance for a Claude blocker is provided by a Codex-help loop that iteratively surfaces the issue, returns a JSON diagnosis, and gives actionable steps to unblock the task.

Can I run iterative debugging sessions that verify outcomes across multiple rounds?

Yes, iterative debugging sessions run up to N rounds, preserving a persistent Codex thread to converge on a resolution and verify outcomes while Claude maintains control over the workflow.

How does Codex CLI integration work for resolving task blockers?

Codex CLI integration provides advisory guidance by capturing analysis, diagnosis, next_steps, missing_context, and open_questions within a strict JSON schema to resolve task blockers.

Do I need a specific schema to capture missing context and open questions during debugging?

A strict JSON schema is required to capture missing_context, open_questions, analysis, and diagnosis, ensuring the Codex assistance returns structured and actionable advisory guidance.

What is the best way to diagnose a workflow blocker with actionable next steps?

The best way to diagnose a workflow blocker is invoking Codex-assisted debugging, which articulates the blocker and provides a structured diagnosis with concrete next steps to unblock work.

Why does Claude need an external advisory loop for unblocking tasks?

Claude needs an external advisory loop because it often encounters blockers during tasks and requires expert, structured guidance to yield a diagnosis, next steps, and verification while remaining in command.