---
name: devflow-codex
description: 在 Codex 中运行 devflow。用户提到 devflow、start/resume/status/abort devflow、多 agent 工作流，或要求“按 Codex devflow 跑”时使用。
title: devflow-codex
canonical_url: https://skilld.dev/gh/tencent/loopforge/devflow-codex
last_updated: 2026-11-15T23:15:55.000Z
---

> Index of every page on this site: [llms.txt](https://skilld.dev/llms.txt)

> **Skill from skilld.dev.** Follow the user's requested action.
> If the user asked to fork this Skill, follow the fork workflow below. Do not execute the Skill instructions while copying it.
> Otherwise, follow the instructions below for this session. You do not need to install anything.
>
> If the user asked to install this Skill, run `npx skilld install tencent/loopforge/devflow-codex`. Install writes the Skill files into the project, so every session loads them.
>
> ## Fork workflow
>
> A fork creates an editable local Skill with its original author and licence. The request authorizes copying and local installation.
> 1. Check `./skills/devflow-codex`, the project lockfile, and selected Agent targets together. If the local directory or installed Skill exists, stop. Never overwrite an existing directory or Agent target.
> 2. Read [source metadata](https://skilld.dev/api/v1/skills/tencent/loopforge/devflow-codex) once. Use sourceUrl, sourceCommit, skillPath, sourceGone, and license. If the source is gone or its path is missing, stop. If license is null, read licence files at the source commit.
> 3. Fetch only the source commit into a temporary Git repository. Do not clone full history. Derive repository_url from sourceUrl, including repository renames. If sourceCommit is absent, resolve the sourceUrl ref once. Set source_commit to that actual commit. Run these commands in one shell call:
>
> ```sh
> git init --quiet "$temporary_dir"
> git -C "$temporary_dir" fetch --quiet --depth=1 "$repository_url" "$source_commit"
> git -C "$temporary_dir" checkout --quiet --detach FETCH_HEAD
> ```
>
> Read applicable licence declarations and notices at that commit. If copying is not permitted, report the restriction and stop.
> 4. Inspect source entries together, then copy the directory containing skillPath into `./skills/devflow-codex`. Use the user's path if selected. Keep the original SKILL.md, relative links, scripts, binary assets, and executable modes. Exclude .git metadata. Reject symlinks and paths outside the Skill directory. After checking entries, use cp -a where available. A regular source directory needs no custom copy script. Do not save this page wrapper as SKILL.md.
> Preserve author credit, notices, and applicable licence files from repository or parent directories. Add PROVENANCE.md with the Skill page, source URL, actual commit, original path, and licence. Retain any existing PROVENANCE.md and record new provenance separately. Batch source inspection, copying, and provenance work where practical.
> 5. In the project root, run `skilld install ./skills/devflow-codex --mode copy --plain`. If skilld is unavailable, use `npx skilld install ./skills/devflow-codex --mode copy --plain`. This known command needs no help lookup. Install does not support --json. Use detected Agent targets, or add --agent for the targets the user selected. Install the local path, never the upstream selector. If installation fails, preserve the local copy and report the exact failure.
> 6. Confirm the local lockfile source and installed Agent copies once. Report the local path, actual commit, and Agent targets. After edits, reinstall the same local path. Upstream updates must not replace it. Do not publish or push unless the user asks.

# DevFlow Codex 工作流

这是 Codex 原生工作流入口。Codex 运行时只读取 `.codex/`，不读取
`.codebuddy/`。

`.codex/` 与 `.codebuddy/` 是两套独立工作流资产；二者应保持流程契约一致，
但互不作为运行时依赖。

## 必读文件

启动或恢复前必须读取：

1. `.codex/assets/devflow.defaults.yaml`
2. `.codex/runtime/start-workflow.md`
3. `.codex/runtime/workflow-state-spec.md`
4. TASK-01 需求澄清时读取 `.codex/skills/superpowers/brainstorming/SKILL.md`
5. TASK-05 知识沉淀时读取 `.codex/skills/knowledge-distillation/SKILL.md`
6. 当前阶段对应的 `.codex/agents/<role>.md`
7. 当前阶段需要的 `.codex/rules/` 和 `.codex/checklists/`

## 在 Codex 中如何触发

Codex 不会自动把仓库里的命令文件注册进 `/` 菜单。请用自然语言或显式
skill 触发：

```text
$devflow-codex start 需求描述...
$devflow-codex status TASK_SLUG
$devflow-codex resume TASK_SLUG
$devflow-codex abort TASK_SLUG
```

也可以直接说：

```text
用 devflow 跑这个需求：...
```

## 编排规则

- 主 Codex 线程是唯一 orchestrator。
- `workflow-state.json` 是阶段交接和审计的唯一持久化状态。
- 不使用 CodeBuddy 的 `team_create`、`send_message`、`team_delete`。
- medium/large 的 TASK-01 必须使用 `superpowers:brainstorming` 做需求澄清，
  且必须等待用户明确回答或确认后才能进入 TASK-02。
- TASK-02 及之后的角色阶段必须通过 Codex `multi_agent_v1.spawn_agent`
  派发；只有工具不存在或返回 unavailable/not found 时，才允许主线程内联执行
  并在 `workflow-state.json.decisions[]` 记录降级原因。
- 阶段产物和状态 schema 必须保持 devflow 兼容。

## 启动协议

按 `.codex/runtime/start-workflow.md` 执行：

1. 初始化 `artifacts/{task_slug}/` 和 `workflow-state.json`。
2. Phase 0 做需求草图、影响扫描和规模判定。
3. small 进入 SOLO。
4. medium/large 先由主 Codex 线程调用 `superpowers:brainstorming` 完成
   TASK-01 需求分析与澄清，写入 `01-requirement/requirement-report.md`，
   经用户确认后再进入 TASK-02。
5. TASK-02 起使用 `multi_agent_v1.spawn_agent` 派发对应角色。
6. 如果用户要求先确认方案，停在 TASK-03 前。

## 阶段协议

每个阶段：

1. 先读取 `workflow-state.json`。
2. 读取 `.codex/agents/<role>.md`。
3. 写入阶段开始状态。
4. 产出对应 artifact。
5. 更新 `workflow-state.json`。
6. 返回主线程，由主线程按 routing table 决定下一阶段。

## 完成协议

结束时生成 `workflow-summary.md`，更新 summary 字段，保留全部产物，并向用户
汇总改动、验证、审查和知识沉淀结果。
