Dual Brain

Dual-Brain

English | 한국어

한국어 안내: 거시적 설계 검토와 미시적 코드 검증을 결합하는 스킬입니다. 한국어 설명과 설치·사용법을 확인하세요.

A portable skill for complementary review and durable project memory. The Right Brain takes the macro view of the problem, architecture, and cross-component effects; the Left Brain takes the micro view of contracts, edge cases, and actual code. The main agent turns both into a finished result.

Dual-Brain is useful for ambiguous designs, risky refactors, unfamiliar integrations, and recurring project decisions. Simple edits use a short local review instead of paying for unnecessary agents. It does not guarantee better results or lower latency on every task.

A clean review is a successful result. Reviewers have no quota for findings or changes. Fixes need current evidence, a concrete consequence, and relevance to the request. Repeated use does not justify speculative hardening, architectural churn, or metadata-only memory writes. Hypotheses stay hypotheses until supported; optional preferences remain unapplied. Stop when the requested work and affected checks are complete.

Supported ways to use it

See platform setup, official sources, and verification limits. These routes do not claim all five products were runtime-tested or expose the same tools. In chat-only mode, the result is analysis or a proposed patch; it cannot claim file writes, tests, independent reviewers, or saved memory that did not occur.

Install

For current Codex local skill discovery, install once at user scope:

Or install for one repository:

Existing installations under ~/.codex/skills/dual-brain that your client already discovers can be updated in place. Avoid installing duplicate copies with the same skill name. Keep SKILL.md and references/ together; the entrypoint loads supporting guidance when…