What problem does it solve? Running OpenAI's Codex CLI effectively requires knowing the right commands, sandbox modes, and approval policies, and integrating it with other AI agents like Antigravity or Claude Code adds further configuration complexity. This Skill provides the operational knowledge to execute Codex CLI non-interactively, delegate sub-tasks, and share skills across agent ecosystems. ## Core Features & Use Cases - Non-Interactive Execution: Run autonomous coding tasks with codex exec using controlled sandbox modes (read-only, workspace-write, danger-full-access) and approval policies. - Automated Code Review: Trigger codex review for uncommitted changes and apply generated patches with codex apply. - Cross-Agent Integration: Configure shared skill directories between Antigravity (.agent/skills) and Codex (~/.codex/skills), plus CODEX.md project instructions. - Use Case: From an Antigravity orchestrator session, delegate a background verification pass by running codex exec -s workspace-write --approve-for-me "Run unit tests and fix failing assertions" against a target repository. ## Quick Start Ask the agent to run a Codex CLI task such as executing codex exec with workspace-write access to implement a feature following a specific skill.
