using-workflows

Compare original and translation side by side

🇺🇸

Original

English

🇨🇳

Translation

Chinese

using-workflows

使用工作流

You pick which recipe the situation needs, fill its args, run it, and keep the closed loop moving. You are not a recipe.

你需要根据场景选择合适的模板，填充参数，运行模板，并保持闭环推进。你本身不是模板。

BYPASS — check FIRST

绕过机制 — 优先检查

One bounded task, single context, no stages, no convergence condition → do it directly (inline, or one worker via

using-tmux-agent-tools

). No recipe, no run dir. When in doubt, bypass — recipes exist for loops, not ceremony.

Red flags that have actually burned us (naming one and proceeding anyway requires a stated reason):

「我記得那支 recipe 內容」— recipes evolve; read the header comment (the args contract) before running. Never guess args.
「先做完再補 run record」— if it meets the run-dir bar below, open it first.
「這個小改不用 gate」— behavior-tier edits to any recipe DO need one consensus-gate round first.

Subagent exemption: delegated workers never enter this router — the dispatcher already routed; workers follow their brief.

如果是单一限定任务、单一上下文、无阶段划分、无收敛条件 → 直接执行（内联执行，或通过

using-tmux-agent-tools

调用单个worker）。无需使用模板，无需运行目录。存疑时选择绕过 — 模板专为循环型工作设计，而非形式化流程。

曾导致问题的警示信号（若明知存在仍继续执行，需说明理由）：

「我记得那个模板的内容」—— 模板会迭代更新；运行前务必阅读头部注释（参数约定）。绝不要猜测参数。
「先做完再补运行记录」—— 如果符合下方运行目录的要求，应先创建运行目录。
「这个小改动不需要校验」—— 对任何模板的行为层级修改，都必须先经过一轮共识校验。

子Agent豁免规则：被委派的worker无需进入此路由 — 调度器已完成路由；worker只需遵循其任务简报。

TRIGGER

触发条件

Loop-shaped work: stages plus a convergence condition. Audit chains, consensus review, plan→build lifecycles, findings triage.

循环型工作：包含多个阶段及收敛条件。例如审计链、共识评审、计划→构建生命周期、问题分类。

From the Continuity trigger (global CLAUDE.md/AGENTS.md)

来自全局CLAUDE.md/AGENTS.md的连续性触发

The global rule routes loop-shaped work here at task start. On arrival:

Match the trigger word (audit / consensus verification / findings triage / root-cause deep-dive / plan→build) to a recipe via SELECT below.
Runtime split: Claude invokes the recipe natively via
```
Workflow
```
; Codex commands it through
```
references/codex-adapter.md
```
(
```
claude-workflow-runner
```
) — the runner is mechanical, so use the cheapest capable model with reasoning effort
```
low
```
.
Write-back duty: every recipe artifact/result lands in the CALLER's
```
.workflow/<YYYYMMDDHHMM>-<slug>/
```
run directory (the one the Continuity rule had you create) — never a detached location. One task = one run dir, shared by convention and recipe alike.

全局规则会在任务启动时将循环型工作路由至此。到达后：

通过下方的SELECT步骤，将触发词（audit / consensus verification / findings triage / root-cause deep-dive / plan→build）匹配到对应模板。
运行时拆分：Claude通过
```
Workflow
```
原生调用模板；Codex通过
```
references/codex-adapter.md
```
（
```
claude-workflow-runner
```
）调用模板 — 运行器是机械性的，因此使用成本最低且具备能力的模型，推理等级设为
```
low
```
。
回写职责：所有模板生成的产物/结果都必须存入调用方的
```
.workflow/<YYYYMMDDHHMM>-<slug>/
```
运行目录（即连续性规则要求你创建的目录）—— 绝不能存入独立位置。一个任务对应一个运行目录，模板需遵循此约定共享该目录。

SELECT

选择模板

Discover live — never recite the recipe list from memory:

bash

ls ~/.claude/workflows/*.workflow.js .claude/workflows/*.workflow.js 2>/dev/null

The inner loop (the ONLY loop — there is no scheduling outer ring; do not invent one):

audit (docs-vs-code | design-vs-code | root-cause-deep-dive)
  → findings-triage        connector ①: askUser → human VERBATIM ·
  │                        briefs → lifecycle · directFix → partitioned run
  → feature-lifecycle-auto thin shell: feature-plan-consensus | plan-pipeline
  │                        → gate ✋ (autoBuild=false: human reads the plan)
  │                        → spec-implement-dual-review-verify
  → re-run the ORIGINATING audit, SAME args
  │                        connector ②: lives in YOU, not in code
  → confirmed == 0 → converged, report · else → back to findings-triage

Entry points off the loop:

weird bug →
```
root-cause-deep-dive-audit
```
· docs/design drifted → the matching audit
ONE artifact needs a second-model verdict →
```
consensus-gate
```
— ONE round, irreversible/behavior-tier changes only; NOT a default station
N-angle generative design consensus →
```
design-consensus
```
"what should this project do next" →
```
project-direction-review
```
recipe fleet inventory / machine drift →
```
workflow-manifest
```

Stage recipes (

feature-plan-consensus

plan-pipeline

spec-implement-dual-review-verify

) are normally reached THROUGH

feature-lifecycle-auto

; call one directly only when you want just that stage. Args auto-fill:

cli

= user's words → repo CLAUDE.md → the COUNTERPART engine (the second brain is always the OTHER engine relative to the current commander: Claude commands → a codex profile; Codex commands → a claude profile; never your own engine, never a hard-coded name) → ask once;

context

= one line (repo abs path + stack + scope). Prefer name invocation over scriptPath.

实时查找模板 — 绝不要凭记忆列举模板列表：

bash

ls ~/.claude/workflows/*.workflow.js .claude/workflows/*.workflow.js 2>/dev/null

核心循环（唯一的循环 — 不存在调度外环；请勿自行创建）：

audit (docs-vs-code | design-vs-code | root-cause-deep-dive)
  → findings-triage        连接器①: askUser → 人工逐字确认 ·
  │                        简报 → 生命周期 · directFix → 分区运行
  → feature-lifecycle-auto 轻量壳层: feature-plan-consensus | plan-pipeline
  │                        → 校验 ✋ (autoBuild=false: 人工审核计划)
  │                        → spec-implement-dual-review-verify
  → 重新运行初始audit，使用相同参数
  │                        连接器②: 内置在你自身中，而非代码里
  → confirmed == 0 → 收敛完成，生成报告 · 否则 → 返回findings-triage

循环外的入口点：

疑难bug →
```
root-cause-deep-dive-audit
```
· 文档/设计偏离 → 匹配对应的audit模板
单个产物需要第二个模型的结论 →
```
consensus-gate
```
— 仅一轮，仅适用于不可逆/行为层级变更；非默认环节
多视角生成式设计共识 →
```
design-consensus
```
“项目下一步该做什么” →
```
project-direction-review
```
模板集群盘点 / 机器漂移检测 →
```
workflow-manifest
```

阶段模板（

feature-plan-consensus

plan-pipeline

spec-implement-dual-review-verify

）通常通过

feature-lifecycle-auto

访问；仅当你只需要该阶段时才直接调用。参数自动填充规则：

cli

= 用户输入 → 仓库CLAUDE.md → 对应引擎（相对于当前指挥者，第二大脑始终是另一个引擎：Claude指挥 → Codex配置文件；Codex指挥 → Claude配置文件；绝不能是自身引擎，绝不能硬编码名称）→ 仅询问一次；

context

= 一行内容（仓库绝对路径 + 栈信息 + 范围）。优先使用名称调用，而非scriptPath。

Cross-runtime execution

跨运行时执行

Both runtimes RUN recipes through this router; only the execution vehicle differs. Claude Code executes

.workflow.js

natively with

Workflow()

. Codex executes via

ADAPTED: claude-workflow-runner

(one bounded Claude runner capsule — mechanics in

references/codex-adapter.md

), under these constraints (Codex-authored, gate-v2 2026-07-19):

Codex MUST freeze recipe name, args, acceptance, author runtime, and reviewer profile before dispatching exactly one Claude runner.
The runner MUST invoke exactly one native
```
Workflow(...)
```
, preserve its return under
```
recipe_result
```
, write schema-v1
```
result.json
```
, then stop.
Any later exception MUST cap nesting at 2, declare child profiles/round ceilings, use unique sessions plus wait-required results, and forbid children from spawning.
Resolve
```
args.cli
```
by the substantive author under review: Codex for Claude-authored work; a non-Codex profile when Codex authored the target; fail closed if unclear.
Human gates MUST return
```
status: paused
```
plus
```
next_action
```
, preserve
```
recipe_result
```
, stop the runner, and resume only via a new explicitly approved invocation.

Evidence MUST say

recipe <name> executed natively on Claude runtime via runner (commanded by Codex)

; adapter or child failure is never recipe PASS.

Runtime matrix (Claude Code =

NATIVE

for all 12; Codex column):

Recipe	Codex
3 audits, `findings-triage` , `design-consensus` , `project-direction-review` , `workflow-manifest` (7, no tmux inside)	`ADAPTED: claude-workflow-runner`
`consensus-gate` (simple verdict outcome only)	`ADAPTED: direct-claude-review` — `references/codex-adapter.md`
`consensus-gate` (as recipe), lifecycle + its 3 stages (5, they launch agent-tmux inside)	`UNAVAILABLE-NATIVE` until nested-runner (depth-2) tests pass — stop and report; do not improvise

两种运行时均通过此路由运行模板；仅执行载体不同。Claude Code通过

Workflow()

原生执行

.workflow.js

。Codex通过

ADAPTED: claude-workflow-runner

执行（一个限定性的Claude运行器胶囊 — 机制详见

references/codex-adapter.md

），需遵循以下约束（Codex编写，gate-v2 2026-07-19）：

Codex必须在调度单个Claude运行器前，冻结模板名称、参数、验收标准、作者运行时和评审者配置文件。
运行器必须精确调用一次原生
```
Workflow(...)
```
，将返回结果保存到
```
recipe_result
```
下，写入schema-v1格式的
```
result.json
```
，然后停止。
任何后续异常必须将嵌套层级限制为2，声明子配置文件/轮次上限，使用唯一会话并要求等待结果，禁止子进程生成。
根据被评审内容的实际作者解析
```
args.cli
```
：Claude编写的内容 → 使用Codex；Codex编写的目标内容 → 使用非Codex配置文件；若不明确则终止执行。
人工校验必须返回
```
status: paused
```
及
```
next_action
```
，保留
```
recipe_result
```
，停止运行器，仅通过新的明确批准调用恢复执行。
证据必须标注
```
recipe <name> executed natively on Claude runtime via runner (commanded by Codex)
```
；适配器或子进程失败绝不能视为模板执行通过。

运行时矩阵（Claude Code = 所有12种模板均为

NATIVE

；Codex列）：

模板	Codex
3种audit、 `findings-triage` 、 `design-consensus` 、 `project-direction-review` 、 `workflow-manifest` （共7种，内部无tmux）	`ADAPTED: claude-workflow-runner`
`consensus-gate` （仅简单结论输出）	`ADAPTED: direct-claude-review` — 详见 `references/codex-adapter.md`
`consensus-gate` （作为模板）、生命周期及其3个阶段（共5种，内部启动agent-tmux）	`UNAVAILABLE-NATIVE` 直到嵌套运行器（深度2）测试通过 — 停止并报告；请勿自行变通

DEFER

延迟规则

Chain recipes from the TOP level only —
```
workflow()
```
nesting cap is 1, and the lifecycle shell spends it.
Run dir (
```
.workflow/<slug>/
```
,
```
codex-dynamic-workflows
```
conventions) ONLY when work spans days, has 2+ phases, or must survive interruption/handoff. Within-chat work: a single results file, or nothing.
Recipe edits: behavior-tier → consensus-gate one round FIRST; wording → direct. Canonical = the agent-scripts repo bundle (
```
skills/using-workflows/workflows/
```
) → redeploy to
```
~/.claude/workflows/
```
; a live edit made machine-side must be folded back into the bundle in the same change.

仅能从顶层链式调用模板 —
```
workflow()
```
嵌套上限为1，且生命周期壳层已占用该额度。
仅当工作跨多天、包含2个以上阶段，或需在中断/交接后保留时，才使用运行目录（
```
.workflow/<slug>/
```
，遵循
```
codex-dynamic-workflows
```
约定）。聊天内工作：只需单个结果文件，或无需文件。
模板编辑：行为层级修改 → 必须先经过一轮共识校验；措辞修改 → 直接编辑。标准版本为agent-scripts仓库包（
```
skills/using-workflows/workflows/
```
）→ 重新部署到
```
~/.claude/workflows/
```
；在机器端进行的实时编辑必须同步合并到包中。

NOT-FOUND

未找到模板

No recipe fits → the work probably is not loop-shaped; bypass. A genuinely new loop shape → propose a new recipe to the user; never improvise a half-recipe inline. Per-recipe reference:

workflows/README.md

(canonical: agent-scripts bundle

skills/using-workflows/workflows/README.md

; deployed copy at

~/.claude/workflows/README.md

无匹配模板 → 该工作很可能不是循环型工作；选择绕过。若确实是新的循环型工作 → 向用户提议新模板；绝不要在内部临时拼凑半完整模板。模板参考文档：

workflows/README.md

（标准版本：agent-scripts包中的

skills/using-workflows/workflows/README.md

；部署副本位于

~/.claude/workflows/README.md

）。