# Codex

> Independent second-opinion workflow for Codex. Use when the user wants an adversarial review, a second-pass challenge, or a separate consultation mode inside the current workflow.

- Skill: `mphaxise/codex` (Agent Skill)
- Install (CLI): `npx skillmds@latest add mphaxise/codex`
- Raw SKILL.md: https://api.skillmd.com/api/skills/mphaxise/codex/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Productivity
- Author: mphaxise (https://skillmd.com/u/mphaxise)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/mphaxise/codex

---


# Codex

Use this skill when the user wants a deliberately independent second opinion instead of a continuation of the same reasoning path.

This port is adapted from `garrytan/gstack` at commit `4d2c8d94d00cc4f4f3d4c26316a4f939ceedc045`.

## Modes

- `REVIEW`: pass/fail second review
- `CHALLENGE`: adversarial attempt to break the current plan or code
- `CONSULT`: focused consultation on a hard question

## Workflow

1. Choose the mode explicitly.
2. Restate the work product or question under review.
3. Re-check it from a different angle, reusing `cross-modal-review` patterns when helpful.
4. Report agreement, disagreement, and confidence clearly.

## Guardrails

- Do not simply repeat the first analysis in softer words.
- Keep the second opinion meaningfully independent.
- If there is no real disagreement, say so plainly.