MCP Server Builder
anthropics/skills
Guides the design and implementation of Model Context Protocol servers in TypeScript or Python, from tool naming and error messages to evaluation.
OC-Flow:为你的 OpenClaw 注入"确定性"灵魂。OC-Flow 完全嵌入在 OpenClaw 体系内,赋予 Agent 完整的流程控制能力:条件分支、循环遍历、精准等待、状态管理。通过 YAML 剧本实现固定流程、多步循环、严苛逻辑的任务。适用场景:财务办公、开发运维、个人助理。
$ npx skills add LeoYeAI/openclaw-master-skills --skill openclaw-workflow -a claude-codeProject install by default; add -g for ~/.claude/skills/.
$ gh skill install LeoYeAI/openclaw-master-skills openclaw-workflow --agent claude-codeProject scope by default; add --scope user for a personal install. Needs GitHub CLI 2.90.0 or later (public preview).
$ git clone --depth 1 https://github.com/LeoYeAI/openclaw-master-skills.git skills-src && mkdir -p .claude/skills && cp -r skills-src/skills/openclaw-workflow .claude/skills/openclaw-workflow && rm -rf skills-srcUse ~/.claude/skills/ instead of .claude/skills for a personal install. The folder must contain SKILL.md.
Claude Code skills documentation · loads skills from .claude/skills/
Install the "openclaw-workflow" agent skill from https://github.com/LeoYeAI/openclaw-master-skills/tree/main/skills/openclaw-workflow into .claude/skills/openclaw-workflow/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "openclaw-workflow", then confirm the skill loads.Claude Code copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.
$skill-installer install https://github.com/LeoYeAI/openclaw-master-skills/tree/main/skills/openclaw-workflowType this inside Codex. $skill-installer <name> installs a curated skill from openai/skills. The installer writes to $CODEX_HOME/skills (default ~/.codex/skills). Restart Codex if the skill does not show up.
$ npx skills add LeoYeAI/openclaw-master-skills --skill openclaw-workflow -a codexProject install goes to .agents/skills/; add -g for ~/.codex/skills/.
$ gh skill install LeoYeAI/openclaw-master-skills openclaw-workflow --agent codexProject scope by default (.agents/skills/); add --scope user for a personal install.
$ git clone --depth 1 https://github.com/LeoYeAI/openclaw-master-skills.git skills-src && mkdir -p .agents/skills && cp -r skills-src/skills/openclaw-workflow .agents/skills/openclaw-workflow && rm -rf skills-srcUse ~/.agents/skills/ instead of .agents/skills for a personal install.
Codex skills documentation · loads skills from .agents/skills/
Install the "openclaw-workflow" agent skill from https://github.com/LeoYeAI/openclaw-master-skills/tree/main/skills/openclaw-workflow into .agents/skills/openclaw-workflow/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "openclaw-workflow", then confirm the skill loads.Codex copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.
$ npx skills add LeoYeAI/openclaw-master-skills --skill openclaw-workflow -a cursorProject install goes to .agents/skills/; add -g for ~/.cursor/skills/.
$ gh skill install LeoYeAI/openclaw-master-skills openclaw-workflow --agent cursorProject scope by default (.agents/skills/); add --scope user for a personal install.
$ git clone --depth 1 https://github.com/LeoYeAI/openclaw-master-skills.git skills-src && mkdir -p .cursor/skills && cp -r skills-src/skills/openclaw-workflow .cursor/skills/openclaw-workflow && rm -rf skills-srcUse ~/.cursor/skills/ instead of .cursor/skills for a personal install.
Cursor skills documentation · loads skills from .cursor/skills/, .agents/skills/, .claude/skills/, .codex/skills/
Install the "openclaw-workflow" agent skill from https://github.com/LeoYeAI/openclaw-master-skills/tree/main/skills/openclaw-workflow into .cursor/skills/openclaw-workflow/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "openclaw-workflow", then confirm the skill loads.Cursor copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.
$ gemini skills install https://github.com/LeoYeAI/openclaw-master-skills.git --path skills/openclaw-workflow--scope user (default) or --scope workspace; --path is the subfolder of the repo that holds the skill; --consent skips the security confirmation prompt.
$ npx skills add LeoYeAI/openclaw-master-skills --skill openclaw-workflow -a gemini-cliProject install goes to .agents/skills/; add -g for ~/.gemini/skills/.
$ gh skill install LeoYeAI/openclaw-master-skills openclaw-workflow --agent gemini-cliProject scope by default (.agents/skills/); add --scope user for a personal install.
$ git clone --depth 1 https://github.com/LeoYeAI/openclaw-master-skills.git skills-src && mkdir -p .gemini/skills && cp -r skills-src/skills/openclaw-workflow .gemini/skills/openclaw-workflow && rm -rf skills-srcUse ~/.gemini/skills/ instead of .gemini/skills for a personal install, then run /skills reload.
Gemini CLI skills documentation · loads skills from .gemini/skills/, .agents/skills/
Install the "openclaw-workflow" agent skill from https://github.com/LeoYeAI/openclaw-master-skills/tree/main/skills/openclaw-workflow into .gemini/skills/openclaw-workflow/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "openclaw-workflow", then confirm the skill loads.Gemini CLI copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.
$ gh skill install LeoYeAI/openclaw-master-skills openclaw-workflowInstalls for Copilot at project scope by default; add --scope user for a personal install. Preview a skill first with gh skill preview. Needs GitHub CLI 2.90.0 or later (public preview).
$ npx skills add LeoYeAI/openclaw-master-skills --skill openclaw-workflow -a github-copilotProject install goes to .agents/skills/; add -g for ~/.copilot/skills/.
$ git clone --depth 1 https://github.com/LeoYeAI/openclaw-master-skills.git skills-src && mkdir -p .github/skills && cp -r skills-src/skills/openclaw-workflow .github/skills/openclaw-workflow && rm -rf skills-srcUse ~/.copilot/skills/ instead of .github/skills for a personal install. Commit .github/skills so cloud agent and code review can use it.
GitHub Copilot skills documentation · loads skills from .github/skills/, .claude/skills/, .agents/skills/
Install the "openclaw-workflow" agent skill from https://github.com/LeoYeAI/openclaw-master-skills/tree/main/skills/openclaw-workflow into .github/skills/openclaw-workflow/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "openclaw-workflow", then confirm the skill loads.GitHub Copilot copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.
$ npx skills add LeoYeAI/openclaw-master-skills --skill openclaw-workflow -a opencodeOpenCode documents no install command of its own. Project install goes to .agents/skills/; add -g for ~/.config/opencode/skills/.
$ gh skill install LeoYeAI/openclaw-master-skills openclaw-workflow --agent opencodeProject scope by default (.agents/skills/); add --scope user for a personal install.
$ git clone --depth 1 https://github.com/LeoYeAI/openclaw-master-skills.git skills-src && mkdir -p .opencode/skills && cp -r skills-src/skills/openclaw-workflow .opencode/skills/openclaw-workflow && rm -rf skills-srcUse ~/.config/opencode/skills/ instead of .opencode/skills for a personal install.
OpenCode skills documentation · loads skills from .opencode/skills/, .claude/skills/, .agents/skills/
Install the "openclaw-workflow" agent skill from https://github.com/LeoYeAI/openclaw-master-skills/tree/main/skills/openclaw-workflow into .opencode/skills/openclaw-workflow/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "openclaw-workflow", then confirm the skill loads.OpenCode copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.
openclaw-workflowOC-Flow:为你的 OpenClaw 注入"确定性"灵魂。OC-Flow 完全嵌入在 OpenClaw 体系内,赋予 Agent 完整的流程控制能力:条件分支、循环遍历、精准等待、状态管理。通过 YAML 剧本实现固定流程、多步循环、严苛逻辑的任务。适用场景:财务办公、开发运维、个人助理。
Openclaw Workflow is an agent skill from LeoYeAI/openclaw-master-skills. OC-Flow:为你的 OpenClaw 注入"确定性"灵魂。OC-Flow 完全嵌入在 OpenClaw 体系内,赋予 Agent 完整的流程控制能力:条件分支、循环遍历、精准等待、状态管理。通过 YAML 剧本实现固定流程、多步循环、严苛逻辑的任务。适用场景:财务办公、开发运维、个人助理。
Its SKILL.md is about 4k tokens, which your agent loads only when the skill is triggered. The skill folder holds 25 other files, including scripts and reference files (for example `README.md`, `README_ZH.md` and `_meta.json`).
It works with Python. The repository describes itself as: 🧠 Curated collection of 1209+ best OpenClaw skills — weekly updated by MyClaw.ai. The licence is MIT.
12 steps, taken from the step headings in SKILL.md.
Read from SKILL.md and the folder at commit e5199b5. It shows what the files ask for, not the result of running them.
Pre-approves nothing: there is no allowed-tools line, so your agent's usual permission prompts apply.
From allowed-tools in the SKILL.md frontmatter.
Ships 5 files in scripts/ (Python, from the files we listed), which the agent can run.
Shell commands in SKILL.md call:
python3From the folder's file list and the shell code blocks in SKILL.md.
No URLs in SKILL.md.
From URLs in SKILL.md, links to its own repository left out.
Names no API keys, tokens, secrets or passwords.
From names ending in _API_KEY, _TOKEN, _SECRET, _KEY or _PASSWORD in SKILL.md.
Openclaw Workflow loads about 4k tokens when it runs, and up to ~12k if it reads all its reference files. Until then it costs about 41 tokens; SKILL.md has 1,120 words of instructions outside code blocks.
Estimates: characters ÷ 4, the usual rule of thumb; real counts depend on the model's tokenizer. Scripts and assets cost tokens only if the agent reads them.
The automated check found no risky patterns in SKILL.md.
Automated static check — not a guarantee. Review scripts before installing. It scans the text of SKILL.md for risky patterns (piping downloads into a shell, reading credential files, hidden Unicode, destructive commands); the scripts in this folder are not scanned.
The full file from LeoYeAI/openclaw-master-skills at commit e5199b5, republished under its MIT licence (© LeoYeAI). 1,120 words, ~3,969 tokens.
.claude/skills/openclaw-workflow/SKILL.md (or your agent's skills folder). This skill also uses 21 other files; get the full folder from GitHub.在不破坏 OpenClaw 灵活性的前提下,按 YAML 剧本执行 100% 确定性逻辑:判断、循环、脚本、LLM 调用、Skill 调用。
这是一个符合 OpenClaw / AgentSkills 目录约定的 Skill:
SKILL.md:触发条件与使用说明scripts/:可执行入口与运行时代码references/:参考文档与示例工作流重要: 脚本内置了 stdout 行缓冲,不需要额外设置
PYTHONUNBUFFERED。
# 运行工作流
python3 {baseDir}/scripts/openclaw_workflow.py execute <workflow.yaml>
# 验证工作流语法
python3 {baseDir}/scripts/openclaw_workflow.py validate <workflow.yaml>
# 列出可用工作流
python3 {baseDir}/scripts/openclaw_workflow.py list
# 查看历史运行
python3 {baseDir}/scripts/openclaw_workflow.py runs
# 从断点恢复
python3 {baseDir}/scripts/openclaw_workflow.py resume <run_id>
# 可视化面板
python3 {baseDir}/scripts/openclaw_workflow.py dashboard推荐目录(OpenClaw 自建/自维护流程):
~/.openclaw/workspace/workflows面板与 CLI 会优先发现以下位置:
{baseDir}/references/examples~/.openclaw/workspace/workflows说明:面板只扫描固定目录,不会递归读取整个 ~/.openclaw/workspace。
示例:
# 运行示例工作流
python3 {baseDir}/scripts/openclaw_workflow.py execute {baseDir}/references/examples/basic_test.yaml
# 运行深度集成测试
python3 {baseDir}/scripts/openclaw_workflow.py execute {baseDir}/references/examples/deep_integration.yaml
# 查看补充设计说明
cat {baseDir}/references/readme.mdJSON 运行记录,包含:
run_id — 运行唯一 IDflow_id — 工作流名称status — success | failed | abortedsteps — 每步执行结果(状态、输出、耗时、重试次数)started_at / finished_at — 起止时间终端同时输出人可读的步骤进度日志。
llm / agent / skill 节点通过 Gateway RPC 调用,不是本地直连模型。name: "我的流程"
steps:
- id: fetch
type: script
command: "curl -s https://api.example.com/data"
- id: analyze
type: llm
prompt: "分析: {{fetch.output}}"
- id: delegate
type: subagent
task: "根据以下分析结果生成一份详细报告: {{analyze.text}}"
label: "报告生成器"
wait: true
timeout: 300
- id: notify
type: message
channel: imessage
target: "+861760051xxxx"
message: "结果: {{delegate.result}}"| 类型 | 说明 | 关键参数 |
|---|---|---|
script | Shell/Python 脚本 | command, inline, timeout |
llm | LLM 推理 (Gateway session) | prompt, thinking, session |
agent | Agent 调用 (Gateway session) | message, thinking, deliver |
subagent | 创建子代理执行独立任务 | task, label, model, wait |
wait_subagents | 等待多个 Subagent 完成并收集结果 | tracker, max_wait, poll_interval |
skill | 调用 OpenClaw Skill (Gateway session) | action, args, instruction |
condition | If-Else 分支 | if, then, else |
loop | 循环遍历 | foreach/times, as, do |
set | 设置变量 | var, value |
log | 日志输出 | message, level |
http | HTTP 请求 | url, method, headers, body |
code | 内联 Python (沙箱) | python |
wait | 延时/等待条件 | seconds, until |
message | 发送消息 | channel, target, message |
{{variable}} — 全局变量{{step_id.output}} / {{step_id.text}} — 步骤输出{{item}} — 循环当前元素{{env.HOME}} — 环境变量每步可配置: retry: 3, retry_delay: 10, on_error: retry|skip|stop
subagent 节点通过 OpenClaw sessions_spawn 工具创建独立子代理。
与 agent 节点的区别:
agent: 在共享 session 中调用主 Agent,同一上下文subagent: 创建独立子代理,有自己的 session、系统提示和模型配置每个 subagent 的创建需要两层 session:
sessions_spawn 工具调用的"载体 session"sessions_spawn 在 Gateway 侧创建的实际子代理 session (agent:main:subagent:<uuid>)sessions_spawn 是 Agent 工具 (不是 Gateway 直接 RPC),所以必须通过 agent_call 在某个 session 中触发。spawn session 本身只是工具调用的容器,子代理真正执行任务的是 child session。
每个 spawn 创建一个临时的 spawn:<hex> session → agent_call 让 Agent 调用 sessions_spawn → 提取 childSessionKey → 立即删除 spawn session。
所有 spawn 复用同一个 factory:<hex> session → 每 20 次轮换新 session → 用 factory_lock 序列化访问。
factory_lock 强制所有 spawn 串行执行,即使在并行循环中原问题 1: 循环中工厂模式强制串行
已通过批量 spawn 解决。循环中自动检测并使用批量创建。
原问题 2: Gmail 24 封邮件超时
批量 spawn 10 个 subagent 仅需 ~77s,24 个预计 ~90-120s,远低于超时限制。
2026-03-18 架构更新: 子会话模式
非循环场景下的单个 subagent 节点不再通过 sessions_spawn 间接创建,而是直接创建新会话执行任务。 会话就是 subagent,省去了 spawn 中间层。
参数:
| 参数 | 必需 | 默认值 | 说明 |
|---|---|---|---|
task | ✅ | — | 子代理要执行的任务描述 |
label | — | (空) | 子代理显示名称 |
model | — | (继承) | 覆盖使用的模型 |
thinking | — | (关) | 思考级别: off/minimal/low/medium/high |
timeout | — | 300 | 超时秒数 |
wait | — | false | 是否等待子代理完成 |
poll_interval | — | 15 | 等待时轮询间隔 (秒) |
mode | — | run | run (一次性) 或 session (持久) |
cleanup | — | auto | keep (保留) 或 auto (wait 完成后自动删除 session) |
throttle_timeout | — | 300 | 并发等待超时 (秒),达到 maxConcurrent 限制时等待空位 |
spawn_timeout | — | 120000 | Gateway 调用超时 (ms) |
spawn_retries | — | 2 | Gateway timeout 重试次数 |
示例 — fire-and-forget:
- type: subagent
task: "整理今天的新闻摘要"
label: "新闻助手"示例 — 等待结果:
- id: research
type: subagent
task: "研究 {{topic}} 并写一份 500 字的分析报告"
label: "研究员"
wait: true
timeout: 600
poll_interval: 20
- type: log
message: "研究结果: {{research.result}}"wait_subagents 节点实现标准的 subagent fan-in (汇合) 模式:
subagent 节点 (wait: false) 创建多个并行子代理wait_subagents 节点轮询所有子代理的 JSONL completion event,全部完成后返回结果列表wait=true 完成后立即删除 child sessionwait_subagents 节点完成后批量清理所有 child session (cleanup: auto)agent:main:openclaw-workflow:<ns>)get_factory_child_session_keys)subagent 节点在创建时会检查两个配置:
agents.defaults.subagents.maxConcurrent (子代理专属限制,默认 20)agents.defaults.maxConcurrent (Gateway 全局嵌入式运行并发限制,默认 4)取两者较小值作为实际限制。并发计数仅统计 :spawn: session (瞬态占位),不统计 :subagent: session (因为完成后仍残留在 sessions.json 中会导致误判)。
subagent 节点对 Gateway timeout 错误内置 2 次重试 (spawn_retries=2),spawn 超时时间 120s (spawn_timeout=120000ms)。
参数:
| 参数 | 必需 | 默认值 | 说明 |
|---|---|---|---|
tracker | ✅ | — | spawn 信息列表 (需包含 spawn_session_key, child_session_key) |
max_wait | — | 600 | 最大等待秒数 |
poll_interval | — | 5 | 轮询间隔秒数 |
extra_fields | — | [] | 从 tracker item 透传到结果中的额外字段名 |
cleanup | — | auto | Session 清理策略: auto (全部清理), completed (仅清理已完成的), keep (不清理) |
示例 — 并行分类:
# 1) 初始化 tracker
- id: init_tracker
type: code
python: "result = []"
# 2) loop 中 fire-and-forget spawn + 收集信息
- type: loop
foreach: "{{items}}"
as: item
do:
- id: spawn_task
type: subagent
task: "分析: {{item.data}}"
wait: false
- id: collect
type: code
python: |
tracker = init_tracker if isinstance(init_tracker, list) else []
tracker.append({
"item_id": item.get("id", ""),
"spawn_session_key": spawn_task.get("spawn_session_key", ""),
"child_session_key": spawn_task.get("child_session_key", ""),
})
result = tracker
# 3) 等待全部完成
- id: wait_all
type: wait_subagents
tracker: "{{init_tracker}}"
max_wait: 600
poll_interval: 5
extra_fields:
- item_id
# 4) 使用结果
- type: agent
message: "汇总: {{wait_all}}"下面补全每个节点的常用写法,优先覆盖实际引擎支持的字段。
| 字段 | 说明 |
|---|---|
id | 步骤唯一标识,建议填写,便于引用 {{step_id.output}} |
name | 人类可读名称,仅用于日志 |
type | 节点类型 |
retry | 失败重试次数 |
retry_delay | 每次重试间隔(秒) |
on_error | retry / skip / stop |
script 节点执行 shell 命令或内联 Python。
支持字段:
command / script / file:三选一inline:内联 Python 代码(会写入临时 .py 执行)timeout:默认 300 秒cwd:工作目录env:环境变量字典- id: fetch_data
type: script
command: "curl -s https://api.example.com/data"
timeout: 30
- id: build_report
type: script
inline: |
import json
print(json.dumps({"ok": True, "ts": "{{env.HOME}}"}, ensure_ascii=False))llm 节点通过 Gateway 调用模型推理,默认复用当前工作流会话。
支持字段:
prompt(必填)thinking:off|minimal|low|medium|hightimeout:秒,默认 120session:shared(默认)或 isolated- id: summarize
type: llm
prompt: "请总结以下内容:{{fetch_data.output}}"
thinking: low
session: sharedagent 节点调用主 Agent。与 llm 相比,支持投递能力(deliver)。
支持字段:
message(必填)thinkingtimeout:默认 300deliver:是否自动投递deliver_channel / deliver_target- id: agent_reply
type: agent
message: "根据 {{summarize.text}} 输出客户可读版本"
deliver: true
deliver_channel: imessage
deliver_target: "+8617600510003"skill 节点让 Agent 在当前会话中调用已有 Skill。
支持字段:
action(必填)args:参数对象instruction:自定义指令(有则优先)timeout:默认 300- id: call_tool_skill
type: skill
action: "transmission.add"
args:
url: "magnet:?xt=..."
instruction: "请添加该任务并返回任务ID"condition 节点条件分支,执行 then 或 else 子步骤。
支持字段:
if(必填):Python 表达式字符串then:条件为真时执行的步骤数组else:条件为假时执行的步骤数组- id: check_items
type: condition
if: "len(load_emails) > 0"
then:
- type: log
message: "有邮件"
else:
- type: log
message: "无邮件"loop 节点循环执行子步骤,支持 foreach 或 times。
支持字段:
foreach:列表/可解析为列表的值times:整数次数(与 foreach 二选一)as / var:循环变量名,默认 itemdo / steps:子步骤数组- id: iterate
type: loop
foreach: "{{items}}"
as: item
do:
- type: log
message: "当前: {{item}}"
- id: retry_three_times
type: loop
times: 3
as: i
do:
- type: log
message: "第 {{i}} 次"set 节点设置全局变量。
支持字段:
var(必填)value(必填)- id: set_topic
type: set
var: topic
value: "PiSugar"log 节点写运行日志。
支持字段:
message(必填)level:默认 INFO- type: log
level: "WARN"
message: "当前数据为空,进入降级路径"http 节点发送 HTTP 请求。
支持字段:
url(必填)method:默认 GETheadersparams:Query 参数body:可为对象/数组/字符串timeout:默认 30- id: request_api
type: http
method: POST
url: "https://api.example.com/v1/report"
headers:
Authorization: "Bearer {{token}}"
body:
title: "日报"
content: "{{summarize.text}}"code 节点执行沙箱 Python。将当前变量与历史步骤输出注入运行环境。
支持字段:
python(必填)结果约定:
result 作为节点输出result,则返回 print 输出文本- id: calc_stats
type: code
python: |
items = load_emails if isinstance(load_emails, list) else []
result = {
"count": len(items),
"subjects": [x.get("subject", "") for x in items[:5] if isinstance(x, dict)]
}wait 节点固定延时或轮询等待条件。
支持字段:
seconds:固定等待秒数until:条件表达式(与 seconds 二选一)poll_interval:默认 5max_wait:默认 300- type: wait
seconds: 10
- type: wait
until: "task_done == True"
poll_interval: 3
max_wait: 180message 节点发送消息。默认通过 Agent 代发,direct: true 时走 CLI 直发。
支持字段:
channel:默认 imessagetarget(必填)message / media:至少一个account:可选账号direct:默认 falsewhen:条件守卫,不满足则跳过- id: notify
type: message
channel: imessage
target: "+8617600510003"
message: "📬 报告如下:{{agent_reply.text}}"
when: "len(agent_reply.text) > 0"subagent 节点创建子代理执行独立任务,支持异步或等待模式。
支持字段:
task(必填)label / model / thinkingtimeout:默认 300wait:默认 falsepoll_interval:默认 5mode:run(默认)/ sessioncleanup:auto(默认)/ keepthrottle_timeout:并发等待超时,默认 300spawn_timeout:Gateway 调用超时 (ms),默认 120000spawn_retries:Gateway timeout 重试次数,默认 2- id: classify_one
type: subagent
task: "请分类此邮件:{{email.body}}"
label: "邮件分类器"
wait: true
timeout: 120
cleanup: autowait_subagents 节点汇合多个 wait: false 子代理,集中等待并收集结果。
支持字段:
tracker(必填)max_wait:默认 600poll_interval:默认 5extra_fields:透传字段cleanup:auto(默认)/ completed / keep- id: wait_all
type: wait_subagents
tracker: "{{init_tracker}}"
extra_fields: [email_id, subject]
cleanup: autoid,避免后续引用困难。 loop + subagent(wait:false) + code(collect) + wait_subagents 是推荐并行模式。Engine 会自动检测此模式并使用批量 spawn(单次 agent_call 并行创建多个 subagent),无需额外配置。 timeout、retry 和 on_error。 llm.session: isolated 或 subagent。 code 节点输出结构化 result。 spawn_batch_size 调整每批数量,默认 8。核心思路: OpenClaw Agent 在单次对话回复中可以并行调用多次 sessions_spawn 工具。利用这一点,engine 在一次 agent_call 中指示 Agent 同时创建 N 个 subagent,而不是每个 subagent 单独调一次。
自动触发条件: _execute_loop() 会自动检测循环体是否符合 "subagent(wait=false) + code(collect)" 模式。如果符合且 len(items) > 1,自动走批量 spawn 路径,无需任何 YAML 配置变更。
批量 spawn 流程:
engine._detect_batch_spawn_pattern(sub_steps) → 检测循环体结构engine._execute_loop_batch_spawn() → 开 factory session → 分批调用 batch_spawn_subagents()nodes.batch_spawn_subagents(items, step_template, ctx, log, bridge, batch_size):batch_size (默认 8) 分批sessions_spawnbridge.agent_call() 发送到 factory sessionbridge.extract_all_spawn_info_from_session_log() 从 JSONL 提取所有 childSessionKey性能对比:
| 模式 | 10 个 subagent | 24 个 subagent (Gmail) | Session 开销 |
|---|---|---|---|
| 子会话模式 (单个) | ~25s | N/A (非循环) | 1 child |
| subagent 串行 | ~250s (25s×10) | ~600s (超时❌) | N + 几个 |
| 批量 spawn | ~77s ✅ | 预计 ~90-120s ✅ | N + 1 factory |
实测数据 (batch_test10, 2026-03-17):
关键代码路径:
engine.py _detect_batch_spawn_pattern() → 模式检测engine.py _execute_loop_batch_spawn() → 批量执行入口nodes.py batch_spawn_subagents() → 核心批量 spawn 逻辑bridge.py extract_all_spawn_info_from_session_log() → 从 JSONL 提取全部 spawn 结果bridge.py 工厂 session 管理 (open/close/rotate/track) — 仍然使用参数:
| 参数 | 默认值 | 说明 |
|---|---|---|
spawn_batch_size | 8 | 每批并行创建的 subagent 数量 (YAML step 级别可配) |
spawn_timeout | 180000ms | 批量 spawn 的 Gateway 超时 (比单个 spawn 的 120000ms 更长) |
spawn_retries | 2 | Gateway timeout 重试次数 |
注意事项:
sessions_spawn 的 cleanup 字段只接受 "delete" 或 "keep",不接受 "auto"min(subagents.maxConcurrent=20, maxConcurrent=8)© LeoYeAI, MIT. Rendered from Markdown: HTML in the file is shown as text, images as links, and headings moved down two levels. Raw file
SKILL.md and 21 other files (scripts, references) in skills/openclaw-workflow of LeoYeAI/openclaw-master-skills.
Open the folder on GitHubat commit e5199b5
Openclaw Workflow next to the 5 skills that share the most tags, products or categories with it. Stars are the repository's; “used in” counts other GitHub owners with a copy.
| Skill | Stars | Used in | Tokens | Auto-check | Licence | Repo updated |
|---|---|---|---|---|---|---|
| Openclaw Workflow this skillLeoYeAI/openclaw-master-skills | 2.2k | — | ~4k | Automated safety check: Pass | MIT | |
| MCP Server Builderanthropics/skills | 180k | 64 repos | ~2.3k | Automated safety check: Pass | Apache-2.0 | |
| PDF Processinganthropics/skills | 180k | 48 repos | ~2k | Automated safety check: Pass | Proprietary | |
| NotebookLM Research AssistantPleasePrompto/notebooklm-skill | 7.8k | 14 repos | ~2.4k | Automated safety check: Notes | MIT | |
| Manim Video Productionbrowser-use/video-use | 28k | 6 repos | ~3k | Automated safety check: Pass | MIT | |
| Code Review ChecklistshareAI-lab/learn-claude-code | 78k | 5 repos | ~1.1k | Automated safety check: Pass | MIT |
anthropics/skills
Guides the design and implementation of Model Context Protocol servers in TypeScript or Python, from tool naming and error messages to evaluation.
anthropics/skills
Handles everyday PDF jobs in Python and on the command line: extract text and tables, merge, split, rotate, watermark, fill forms, encrypt and OCR.
PleasePrompto/notebooklm-skill
Lets Claude Code ask questions of your Google NotebookLM notebooks through browser automation and return answers grounded in your uploaded sources.
browser-use/video-use
Produces math and technical explainer videos with Manim Community Edition: concept animations, equation derivations, algorithm walkthroughs and data stories.
shareAI-lab/learn-claude-code
Reviews code against a five-part checklist covering security, correctness, performance, maintainability and testing, and reports findings in a fixed format.
hugohe3/ppt-master
Generates editable PowerPoint decks, rebuilds slides from images, fills .pptx templates and polishes existing presentations through routed workflows.
LeoYeAI/openclaw-master-skills
Manages pipelines on a DevOps quality and efficiency platform through its OpenAPI: list workspaces and templates, create, update, run and cancel pipelines, and read run records.
LeoYeAI/openclaw-master-skills
Patches OpenClaw's Feishu extension so an edited document triggers an isolated agent session that reads the doc and replies inline, turning it into a live chat space.
LeoYeAI/openclaw-master-skills
Multi-context memory management system for OpenClaw agents with group-isolated storage, global shared memory, workspace organization, and group-specific skills isolation.
LeoYeAI/openclaw-master-skills
Runs a brand's AI-search visibility work end to end: diagnosing how AI platforms represent it, repositioning it, producing AI-optimized content and monitoring ongoing mentions.
LeoYeAI/openclaw-master-skills
Installs and authenticates the gws CLI, then automates Gmail, Drive, Sheets, Calendar, Docs, Chat and Tasks with ready-made recipes, persona bundles and security audits.
LeoYeAI/openclaw-master-skills
Runs four advisor roles, a fitness coach, nutritionist, data analyst and TCM practitioner, to build a health profile and track workouts, diet and wellness over time.
Works with
OC-Flow:为你的 OpenClaw 注入"确定性"灵魂。OC-Flow 完全嵌入在 OpenClaw 体系内,赋予 Agent 完整的流程控制能力:条件分支、循环遍历、精准等待、状态管理。通过 YAML 剧本实现固定流程、多步循环、严苛逻辑的任务。适用场景:财务办公、开发运维、个人助理。. Openclaw Workflow is an agent skill from LeoYeAI/openclaw-master-skills.
Run `npx skills add LeoYeAI/openclaw-master-skills --skill openclaw-workflow -a claude-code`. Or copy the skill folder (skills/openclaw-workflow in LeoYeAI/openclaw-master-skills) into .claude/skills/openclaw-workflow in your project. Claude Code loads it when a task matches its description.
Run `npx skills add LeoYeAI/openclaw-master-skills --skill openclaw-workflow -a codex`. Or copy the skill folder (skills/openclaw-workflow in LeoYeAI/openclaw-master-skills) into .agents/skills/openclaw-workflow in your project. Codex loads it when a task matches its description.
Cursor, Gemini CLI, GitHub Copilot and OpenCode also load SKILL.md folders. With the skills CLI, run `npx skills add LeoYeAI/openclaw-master-skills --skill openclaw-workflow -a cursor` (or -a gemini-cli, github-copilot or opencode for the others). To copy it by hand, put the folder in .cursor/skills/openclaw-workflow, .gemini/skills/openclaw-workflow, .github/skills/openclaw-workflow and .opencode/skills/openclaw-workflow in your project.
Going by SKILL.md and its folder, Openclaw Workflow needs Python for the scripts in its folder and the command-line tools its instructions call (python3). Our summary lists: Python 3.
SKILL.md contains no URLs. Any network use would come from the scripts or tools the agent runs. This is read from the text; nothing was executed.
Our automated static check of SKILL.md found no risky patterns, such as piping downloads into a shell, reading credential files or hidden Unicode. It is not a guarantee. The check reads SKILL.md only: the scripts in the folder are not scanned, so read them before running anything.
Openclaw Workflow is published under the MIT licence (the repository's licence). It allows redistribution, so the full SKILL.md is shown on this page.
About 4k tokens (SKILL.md is roughly 16k characters). Agents keep only the skill's name and description in context until a task matches; then they load SKILL.md in full. Its references folder adds about 7.8k tokens, read only when the agent opens those files.
Skills that share tags, products or a category with Openclaw Workflow: MCP Server Builder (anthropics/skills, 180k stars), PDF Processing (anthropics/skills, 180k stars), NotebookLM Research Assistant (PleasePrompto/notebooklm-skill, 7.8k stars) and Manim Video Production (browser-use/video-use, 28k stars). The comparison table on this page puts their stars, adoption, token cost, safety result and licence side by side.
LeoYeAI (a GitHub user) maintains it in LeoYeAI/openclaw-master-skills, which has 2,159 GitHub stars. The repository holds 972 skills in this directory. The repository was last updated on July 20, 2026.
Source: LeoYeAI/openclaw-master-skills on GitHub. Facts on this page come from the repository at the commit we read; the author's words are quoted as theirs.