codex-orchestrator-brief

仓库创建 2026年4月30日最近提交 21 天前SkillHot 收录 20 天前

▸ 精选理由

把人工决策转为可供模型执行的规范，适合自动化交付场景

▸ 风险提示

需要访问仓库代码，可能要求敏感权限

这个 Skill 做什么

分析仓库并产生供实现模型自动运行的完整移交包与实现规范

分析仓库并产出给实现模型（如 Codex）能直接跑的交接包：含证据驱动的 refactor-instructions.md、关键预实施问题、以及带 implementer/verifier 子会话和 waves 的 orchestrator prompt、评分细则与 PR 合并策略。用于把实现工作交给自动化模型时，确保执行步骤、校验点和合并规则都写清，方便模型自治与验收。

▸ 展开 SKILL.md 英文原文

Analyze a repository and produce a complete handoff package for an implementation model (Codex, Opus, etc.) that will run autonomously — an evidence-based refactor-instructions.md spec, a short list of pre-implementation questions, and a /goal-based orchestrator prompt with implementer/verifier subsessions, waves, a checkable rubric, PR merge policy, and a cross-wave memory log. Use whenever the user wants to hand work off to codex or another model via written instructions — phrases like "codex に渡す指示書", "リファクタ指示書を作って", "orchestrator 用のプロンプト", "/goal で完遂させたい", "実装担当モデルへの handoff", "wave で実行させたい", or asks to design implementer/verifier loops. Do NOT use when the user wants Claude to drive codex sessions directly itself (that is codex-tdd-orchestration), wants the refactoring done in this session, or wants an existing prompt/system-prompt improved or optimized (that is prompt-optimizer — this skill only authors new handoff specs from repo analysis).

Skill 开发管理移交包实现指令证据化Codex

Stars

Forks

仓库内 Skill

7 日增星

安装 / 使用

给你的 Agent 一句话（通用）

帮我安装这个 skill：https://raw.githubusercontent.com/Eotel/skills/main/codex-orchestrator-brief/SKILL.md

或 curl 直取 SKILL.md

curl -fsSL "https://raw.githubusercontent.com/Eotel/skills/main/codex-orchestrator-brief/SKILL.md"

SKILL.MD 节选查看完整文件 ↗

# Codex Orchestrator Brief

> **Load the shared core first.** Invoke `codex-orchestration-core` via the Skill
> tool and read its orchestration-loop invariants (numbered 1–11) before applying
> this skill. They are the single source of truth for the loop's philosophy; this
> skill adds only what is unique to *authoring a written handoff package*. Where a
> step below cites "core #N", that invariant lives in the core skill, not here.

This skill produces the package that lets an implementation model complete a
repo-wide refactor (or similar multi-wave task) **without the author in the loop**.
You write; another model runs. The package has three deliverables:

1. `refactor-instructions.md` — t

via SKILL·HOT · 数据来自 GitHub 公开信息 · 原文版权归作者所有