codex

name	codex
description	Use when the user asks to run Codex CLI (codex exec, codex resume) or references OpenAI Codex for code analysis, code review, or automated editing

Codex Skill Guide

Before running Codex, follow the shared logging and summary protocol in ../_shared/agent-cli.md.

Running a Task

1. Ask the user (via the current host's user-question or approval mechanism) which profile, model, and reasoning effort to run in a single prompt.
   • Profile flag: use -p <profile_name> only when the user requests a non-default profile. When omitted, Codex uses the default profile.
   • Default profile: use gpt-5.4 with high or gpt-5.5 with high. Default to gpt-5.3-codex if the user has no model preference.
   • -p llm_router: only use deepseek-v4-pro with xhigh.
   • -p aihubmix: only use gpt-5.4 with high or gpt-5.5 with high.
2. Select the sandbox mode required for the task. Default to --sandbox read-only. For implementation/editing tasks, first use superpowers:using-git-worktrees to create a fresh git worktree, then run Codex there with --sandbox workspace-write.
3. Assemble the command with the appropriate options:
   • -p, --profile <profile_name>
   • -m, --model <MODEL>
   • --config model_reasoning_effort="<xhigh|high|medium>"
   • --sandbox <read-only|workspace-write|danger-full-access>
   • --full-auto
   • -C, --cd <DIR>
   • --skip-git-repo-check
4. Always use --skip-git-repo-check.
5. When continuing a previous session, use codex exec --skip-git-repo-check resume --last via stdin. When resuming don't use any configuration flags unless explicitly requested by the user e.g. if he specifies the profile, model, or reasoning effort when requesting to resume a session. Resume syntax: echo "your prompt here" | codex exec --skip-git-repo-check resume --last. All flags have to be inserted between exec and resume.
6. When reviewing uncommitted changes, use (echo "Review the following uncommitted diff."; git diff) | codex exec --skip-git-repo-check - and other flags as needed between exec and --skip-git-repo-check.
7. Do not suppress stderr. Capture stdout and stderr into full.md, and require the prompt to write summary.md.
8. Run the command, inspect the full log and summary, and summarize the outcome for the user.
9. After Codex completes, inform the user: "You can resume this Codex session at any time by saying 'codex resume' or asking me to continue with additional analysis or changes."

Quick Reference

Use case	Sandbox mode	Key flags
Read-only review or analysis	read-only	--sandbox read-only
Apply edits in fresh worktree only	workspace-write	--sandbox workspace-write --full-auto
Permit network or broad access in fresh worktree only	danger-full-access	--sandbox danger-full-access --full-auto
Resume recent session	Inherited from original	echo "prompt" | codex exec --skip-git-repo-check resume --last (no flags allowed)
Run from another directory	Match task needs	-C <DIR> plus other flags
Use a named profile	Match task needs	-p <profile_name> before model/sandbox flags
llm_router profile	Match task needs	-p llm_router -m deepseek-v4-pro --config model_reasoning_effort="xhigh"
aihubmix profile	Match task needs	`-p aihubmix -m <gpt-5.4

Following Up

• After every codex command, immediately use the current host's user-question or approval mechanism to confirm next steps, collect clarifications, or decide whether to resume with codex exec resume --last.
• When resuming, pipe the new prompt via stdin: echo "new prompt" | codex exec resume --last. The resumed session automatically uses the same profile, model, reasoning effort, and sandbox mode from the original session.
• Restate the chosen profile, model, reasoning effort, and sandbox mode when proposing follow-up actions.

Error Handling

• Stop and report failures whenever codex --version or a codex exec command exits non-zero; request direction before retrying.
• Before you use high-impact flags (--full-auto, --sandbox workspace-write, --sandbox danger-full-access, --skip-git-repo-check) ask the user for permission using the current host's user-question or approval mechanism unless it was already given. Write-capable sandbox flags require a fresh git worktree.
• When output includes warnings or partial results, summarize them and ask how to adjust using the current host's user-question or approval mechanism.