# Cross Modal Review

> GBrain-inspired second-pass review for Codex. Use when the user wants another perspective, a quality gate, or a deliberate re-check before finalizing work.

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

---


# Cross-Modal Review

Use this skill when the user wants a second-pass review of a draft, decision, or artifact before finalizing it.

This port is adapted from `garrytan/gbrain` at commit `0b47afbf402a4e27a648bb9d131ce584461461ea`.

## Important Adaptation

Upstream GBrain assumes an explicit second-model routing chain. In Codex, the equivalent is a deliberate second-pass review against the originating workflow's contract, often using:

- the existing `review` skill
- a separate verification pass
- a restated quality bar before final output

## Workflow

1. Capture the work to review.
2. Restate the contract or quality bar it was meant to satisfy.
3. Re-check the work from a different angle:
   - correctness
   - missing edge cases
   - contract fit
   - clarity
4. Report agreement and disagreement explicitly.

## Current Upstream Coverage

Invoke the second modality only when it adds independent evidence or adversarial value. Support standard artifact review, code-review handoff, and explicit challenge mode; route refusals or unavailable model paths honestly and preserve the user-sovereignty rule for every recommendation.

## Guardrails

- Do not auto-apply review changes without deciding them explicitly.
- Review against promises and requirements, not just vague taste.
- Keep the second pass meaningfully different from the first.

