codex
Use when the user asks to run Codex CLI (codex exec, codex resume) or references OpenAI Codex for code analysis, refactoring, or automated editing
Security Assessment
About codex
The Codex skill enables users to run Codex CLI commands for code analysis, refactoring, and automated editing. It integrates with OpenAI's Codex, allowing users to run tasks like executing commands (`codex exec`), resuming previous sessions (`codex exec resume`), and analyzing code with specific reasoning efforts. This skill simplifies interaction with Codex, providing users with tailored options to perform various code-related tasks efficiently. It offers a versatile solution for automating code review processes, refactoring code, or editing projects with minimal user intervention.
Main features include model selection (gpt-5 or gpt-5-codex), reasoning effort configuration (low, medium, high), and sandbox mode management for different levels of access control (read-only, workspace-write, and full access). Users can also resume previous sessions or run tasks from specific directories. The skill provides flexibility in selecting task-specific configurations and enhances workflow automation. With simple commands, users can review code, apply edits, or allow for broader network access as needed, with full transparency in error reporting and session handling.
This skill is ideal for developers, software engineers, and anyone working on code analysis or automation tasks. It is especially useful for those who need to analyze or refactor large codebases, apply automated edits, or need quick code reviews without manual intervention. The Codex skill can be used for various programming languages and project types, offering a streamlined, automated approach to improve coding efficiency and productivity.
FAQ
How do I choose the reasoning effort for a Codex task?
When running a task, you will be asked to select a reasoning effort: low, medium, or high. This determines the level of complexity and thoroughness Codex uses when analyzing or editing your code.
Can I resume a previous Codex session?
Yes, you can resume a previous session using the command `echo 'your prompt' | codex exec resume --last`. The resumed session will inherit the model, reasoning effort, and sandbox mode from the original run.
What are the different sandbox modes in Codex?
Codex offers three sandbox modes: read-only, workspace-write, and danger-full-access. The default is read-only, but you can switch to workspace-write for local edits or danger-full-access for broader network permissions.
What happens if a Codex command fails?
If a `codex exec` command fails or returns an error, the process will stop, and you will be prompted for further direction before retrying the task.
Is network access allowed during Codex tasks?
Yes, but only if you explicitly choose the `danger-full-access` sandbox mode. Be sure to confirm with the user before selecting this mode, as it provides broad access to the system.
Install codex
Quick Setup:
- Copy the skill folder to
.claude/skills/ - Claude will automatically detect and use the skill
Repository
bear2u/my-skills