What problem does it solve? Delegating autonomous coding work to OpenAI's Codex CLI requires knowing its interactive terminal behavior, sandbox modes, and git repository constraints. This Skill provides the exact invocation patterns, flags, and monitoring workflows to run Codex reliably from an agent terminal. ## Core Features & Use Cases - One-Shot and Background Execution: Run codex exec for single tasks or launch long-running jobs in background mode and monitor them with process polling and logging. - PR Reviews and Batch Issue Fixing: Clone repositories to temp directories for safe reviews, and use git worktrees to fix multiple issues in parallel with separate Codex processes. - Sandbox and Gateway Handling: Choose between --full-auto, --yolo, and --sandbox danger-full-access modes, including workarounds for bubblewrap failures in gateway/service contexts. - Use Case: You need to fix issues #78 and #99 simultaneously. Create two git worktrees, launch a background Codex process in each with a targeted prompt, monitor progress, then push branches and open PRs with gh. ## Quick Start Ask the agent to run Codex in one-shot mode inside your git repository, for example by requesting it to execute codex exec with a prompt like adding a dark mode toggle to the settings page.