How does the Codex validation process work?

It utilizes the mcp__codex__spawn_agent tool to run a background review of Claude's proposed plans or implementation against specific constraints and best practices.

Is plan validation mandatory for every task?

No, plan validation is optional and typically reserved for complex tasks or executed only when the user explicitly requests or agrees to a verification step.

When should I use the Guardrail Loop skill?

Use it for non-trivial tasks involving more than three steps, significant architectural changes, or multiple file modifications where code quality is critical.

What types of issues can this skill detect?

The loop identifies logic errors, missing edge cases, type safety violations (like null/undefined handling), performance bottlenecks, and security vulnerabilities.

How are the validation results presented?

Results are summarized into three priority levels: Critical (must fix), Warnings (recommended improvements), and Suggestions (minor optimizations).

Claude Codex Guardrail Loop

Name: Claude Codex Guardrail Loop
Author: munlucky

bymunlucky

0•

보안 및 테스팅

Automates a multi-stage quality gate loop using Codex for rigorous planning validation and implementation review.

This skill establishes a high-integrity engineering loop between Claude and Codex to ensure code quality and architectural consistency. It implements a structured 'Plan-Validate-Implement-Review' workflow that acts as a quality gate for non-trivial tasks, such as major refactors or new feature development. By leveraging Gemini-inspired prompt strategies and explicit guardrails, it identifies logic errors, type safety issues, and edge cases before and after code changes are applied, ensuring that your AI-driven development remains safe, predictable, and compliant with project conventions.

주요 기능

01Automated background reviews using the mcp__codex__spawn_agent tool

02Standardized Plan-Implementation-Review reporting format

030 GitHub stars

04Multi-stage quality gates for both planning and implementation phases

05Constraint-based guardrails to prevent unauthorized dependency or file changes

06Categorized issue prioritization including Critical, Warning, and Suggestions

사용 사례

01Ensuring strict adherence to project-specific conventions and type safety during implementation

02Validating complex architectural changes and multi-file refactors before code is written

03Performing a final quality check and edge-case review before build or deployment

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add munlucky/claude-command-relay claude-codex-guardrail-loop

For use in Claude.ai and ChatGPT

Download Skill