Codex in Claude

Run Codex from Claude Code for adversarial code review and task delegation.

Skills

• codex-in-claude:adversarial-review: Invokes Codex as an adversarial reviewer running at the user's configured sandbox and permission defaults, has Claude triage the feedback, applies accepted in-scope fixes for current working-tree changes, validates the result, and summarizes the outcome.
• codex-in-claude:using-codex-cli: Internal contract for non-interactive Codex CLI use — command recipes, sandbox modes, sessions, structured output, and GPT-5.5 prompting guidance. The review skill and the codex agent use it; it is not meant for direct invocation.

Agents

• codex: A general-purpose proxy agent that forwards a delegated task to codex exec and returns the result. Use it to bring Codex in as a teammate for implementation, diagnosis, or research tasks, or as a subagent in multi-agent workflows. Review requests belong to the adversarial-review skill instead.

Requirements

• codex must be available on PATH.
• Codex must already be installed, authenticated, and ready to run non-interactively.
• Codex needs network access and writes to ~/.codex. Run codex commands outside any calling sandbox, for example excluded from Claude Code's Bash sandbox. Codex's own OS-level sandbox is the enforcement boundary.

Claude Code support

• Claude Code loads this plugin from .claude-plugin/plugin.json; the root plugin.json is the Agent Plugins 1.0.0 manifest that other clients read.
• Skill content lives in skills/ and agent definitions in agents/ at the plugin root.