codex-skill

Automate code implementation with Codex CLI across sandbox modes.

1.6k|244|Updated Jul 18, 2025
One-click install
npx skills add https://github.com/feiskyer/claude-code-settings --skill codex-skill
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: codex-skill
Source: https://github.com/feiskyer/claude-code-settings/tree/main/plugins/codex-skill/skills/codex-skill
Command: npx skills add https://github.com/feiskyer/claude-code-settings --skill codex-skill

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

This Skill automates coding tasks, feature implementation, and plan execution without constant manual approval, saving developers significant time and reducing repetitive work.

Core Features & Use Cases

  • Autonomous Execution: Completes tasks from start to finish without seeking approval for each action, making confident decisions.
  • Multiple Operating Modes: Supports read-only for analysis, workspace-write for programming, and danger-full-access for system-level operations.
  • Model Selection: Specify which model to use (e.g., gpt-5, gpt-5.1) for tailored performance and capabilities.
  • Use Case: Use this skill to automatically refactor the payment processing module using gpt-5.1 in workspace-write mode, then output a JSON report of changes.

Quick Start

Use the codex-skill to implement the user authentication feature with full-auto mode.

Frequently Asked Questions about codex-skill

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

FAQPage Schema
How do I automate code refactoring without manual approval for each step?

Automate code refactoring by running codex-skill in workspace-write mode, which executes tasks autonomously from start to finish without seeking approval. Specify your model (e.g., gpt-5.1) and the skill handles scaffolding, feature development, bug fixes, and refactoring end-to-end, then outputs a JSON summary of all changes.

Can I use autonomous code generation for feature implementation across different sandbox modes?

Yes, codex-skill supports three sandbox modes: read-only for safe analysis, workspace-write for programming tasks, and danger-full-access for system-level operations. Choose the mode matching your task scope to implement features, fix bugs, or install dependencies with full autonomy.

What's the best way to automate task execution without constant developer intervention?

Use codex-skill's hands-off execution to complete coding tasks from start to finish with confident autonomous decisions. It handles model selection, structured JSON output, progress streaming, and generates final summaries, eliminating repetitive manual approvals throughout the workflow.

How does model selection affect code generation and task execution?

Model selection (e.g., gpt-5, gpt-5.1) tailors performance and capabilities to your specific coding task. Codex-skill lets you specify which model to use, optimizing for your requirements—whether prioritizing speed, code quality, or specialized refactoring and feature development.

Can I get structured output and progress tracking when automating code tasks?

Yes, codex-skill provides progress streaming during execution and outputs structured JSON reports with final summaries. This gives you real-time visibility into autonomous code generation and a complete record of changes, dependency installations, and refactoring applied.

When should I use autonomous code execution instead of manual development?

Use autonomous execution for repetitive coding tasks like scaffolding, refactoring, dependency management, and bug fixes where manual approval at each step creates friction. It's ideal when you need confident hands-off implementation without constant intervention while maintaining control through mode selection and model choice.