Codex CLI

What it does

Codex CLI skill provides structured guidance for delegating coding tasks to an OpenAI Codex terminal agent. It defines command patterns, sandbox modes (read-only, workspace-write, danger-full-access), recommended flags, and error handling. The skill is intended for non-interactive execution via codex exec, including session resume/fork semantics and model/effort defaults. It helps agents run safe analysis, apply edits with --full-auto, and capture structured JSON output when needed.

When to use it

Trigger this skill when the user explicitly requests Codex, wants a second AI opinion, needs Codex-specific features (sandboxed execution, session resume/fork), or asks to run/resume/fork a Codex session. Do NOT use this skill for general coding tasks that a different agent (e.g., Claude) can handle directly.

What's included

• Scripts: none detected, but detailed command examples and flag tables are present in the skill body.
• References: attribution to an upstream codex skill and model/pricing notes.
• Instructions: clear command patterns for different task types, sandbox guidance, session resume examples, model table, flags, and error-handling policies.

Compatible agents

Best for CLI-enabled agents and environments that can run the codex binary or otherwise orchestrate Codex sessions (CLI agents, Copilot/Codex-integrated tools).