codex

Invokes OpenAI Codex CLI with xhigh reasoning to review code, plans, options, and ideas.

28|7|Updated Mar 30, 2026
One-click install
npx skills add https://github.com/DK625/OneHammer --skill codex-dk625
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: codex
Source: https://github.com/DK625/OneHammer/tree/main/.claude/skills/codex
Command: npx skills add https://github.com/DK625/OneHammer --skill codex-dk625

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Getting a reliable second opinion on code, plans, or architectural options usually requires copy-pasting context into another tool and manually configuring the right model flags. This Skill invokes the OpenAI Codex CLI (GPT-5.4, xhigh reasoning) directly from the terminal with the correct subcommands, explicit model flags, and MCP servers disabled to prevent lost responses on long reasoning sessions. ## Core Features & Use Cases - Code Review via Git: Runs codex exec review against uncommitted changes, a specific commit, or a base branch, with optional custom focus prompts. - Options, Idea, and Plan Review: Builds self-contained heredoc prompts from conversation context so Codex can evaluate trade-offs, risks, and sequencing. - Reliability Safeguards: Forces -m gpt-5.4 and model_reasoning_effort=xhigh regardless of user config, and disables configured MCP servers to avoid mid-run disconnections. - Use Case: After drafting an execution plan for a new feature, type /codex review plan to have GPT-5.4 critique completeness, sequencing, and risks before implementation begins. ## Quick Start Ask the assistant to run /codex review code to have Codex review your uncommitted changes with xhigh reasoning.

Frequently Asked Questions about codex

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I review uncommitted code changes with Codex CLI?

Run codex exec review --uncommitted with explicit flags -m gpt-5.4 and -c model_reasoning_effort=xhigh. The review subcommand automatically diffs against git, and you can append a prompt describing the code's intent for more accurate feedback.

What is the difference between codex exec and codex exec review?

codex exec review is the dedicated subcommand for git-based code review, supporting flags like --uncommitted, --commit, and --base. Plain codex exec is used for freeform prompts such as reviewing plans, options, or ideas passed via heredoc.

Why should MCP servers be disabled when running Codex with xhigh reasoning?

Codex sessions with xhigh reasoning run 60-180 seconds or longer, and MCP connections frequently disconnect mid-run, causing the entire response to be lost. Disable each configured server with -c mcp_servers.<name>.enabled=false, referencing only servers that exist in ~/.codex/config.toml.

Can Codex review plans or ideas instead of code?

Yes. Pass the plan, options, or idea in a heredoc prompt to codex exec with the model and reasoning flags. Include context such as relevant file paths or requirements, and ask Codex to evaluate trade-offs, risks, sequencing, and missing steps.

Why does Codex review fail with an invalid transport error?

This happens when a command references an MCP server that does not exist in ~/.codex/config.toml. Check the [mcp_servers] section of the config and only disable servers that are actually configured.