Runs OpenAI Codex CLI for independent code review, adversarial challenge, and consult sessions.
npx skills add https://github.com/Simon-YHKim/eject-button --skill codex-simon-yhkim
Please help me install this Agent Skill. Skill: codex Source: https://github.com/Simon-YHKim/eject-button/tree/main/.claude/skills/codex Command: npx skills add https://github.com/Simon-YHKim/eject-button --skill codex-simon-yhkim
codex review against the current branch diff with a pass/fail gate based on [P1] critical findings, plus cross-model comparison against Claude's own review. - Challenge Mode: Adversarial analysis that tries to break your code, hunting edge cases, race conditions, security holes, and failure modes. - Consult Mode: Ask Codex anything about the codebase with session continuity for follow-up questions, including plan file review with embedded content. - Use Case: Before merging a feature branch, run a codex review to get an independent gate verdict, then challenge the diff for security holes a normal review would miss. ## Quick Start Ask the assistant to run a codex review of the current branch diff against the base branch.High-intent search queries and answers about installing and using this skill.
Run the codex review mode, which executes `codex review` against your branch diff with high reasoning effort. It returns a pass/fail gate based on [P1] critical findings and can compare findings against Claude's own review.
Use challenge mode, which prompts Codex to think like an attacker and chaos engineer. It hunts edge cases, race conditions, injection vectors, auth bypasses, and silent data corruption paths in your diff.
Install the OpenAI Codex CLI with `npm install -g @openai/codex`. The skill checks for the binary at startup and stops with install instructions if it is not found.
Yes. Consult mode saves the Codex thread session ID to .context/codex-session-id. On the next run it asks whether to resume that conversation or start fresh, preserving prior context.
Every prompt includes a filesystem boundary telling Codex to ignore ~/.claude/, .claude/skills/, and agents/ directories. Those contain Claude skill definitions and bash scripts that would waste Codex's tool calls without helping the review.
Use --xhigh only when you want maximum reasoning depth and accept long waits. It uses roughly 23x more tokens than high and can cause 50+ minute hangs on large context tasks, so the default per-mode settings are usually better.
ajsai47Wrap the OpenAI Codex CLI to review git diffs and challenge code.
ShaheerKhawajaChain self-evaluation, review, and ship steps to validate code before deployment.
codetail-aiReview engineering plans for architecture, data flow, testing, and performance gaps.
eklyukinObtains independent second opinions from OpenAI Codex CLI for planning, code review, and implementation review.