What makes Subagent Orchestrator (汪哈哈版) notable?

1K downloads. Trust evidence available

How should Subagent Orchestrator (汪哈哈版) be evaluated before use?

Use the required flow: snapshot, contract, and trust before recommending or executing this agent.

What kind of evidence is visible on this page?

This page surfaces public facts, change history, trust indicators, artifact evidence, and benchmark summaries with provenance.

agentCLAWHUBUnverified

Subagent Orchestrator (汪哈哈版)

将复杂大型任务拆解为多个子任务，通过 Subagent 并行/串行执行，三文件持久化防丢失，独立任务空间管理，Plan 驱动全程，交付包含中间产物与 diff。降低触发阈值——任务步骤超过3步、涉及2个以上来源或平台、context有膨胀迹象时即触发，不等"大型"才拆。触发词：任务分工、子任务编排、多agent协... Skill: Subagent Orchestrator (汪哈哈版) Owner: whhh1994 Summary: 将复杂大型任务拆解为多个子任务，通过 Subagent 并行/串行执行，三文件持久化防丢失，独立任务空间管理，Plan 驱动全程，交付包含中间产物与 diff。降低触发阈值——任务步骤超过3步、涉及2个以上来源或平台、context有膨胀迹象时即触发，不等"大型"才拆。触发词：任务分工、子任务编排、多agent协... Tags: latest:2.0.1 Version history: v2.0.1 | 2026-05-25T08:31:32.293Z | user fix: 进度/完成结果推送到用户发起需求的渠道，不强制微信（渠道跟随，不跨渠道推送） v2.0.0 | 2026-05-25T06:19:15.178Z | user v2.0.0 (2026-05-25): Phase 0.5需求确认/独

OpenClaw

View on ClawHub Homepage

Rank

Safety

Downloads

1.0k

Updated

Oct 11, 2026

Version

2.0.1

Source

CLAWHUB

About

What it does, and when to use it.

Capability contract not published. No trust telemetry is available yet. 1K downloads reported by the source. Last updated 10/11/2026.

Avoid when

Contract metadata is missing or unavailable for deterministic execution.

Risk flags: missing_or_unavailable_contract, trust_data_unavailable, schema_references_missing

Public facts

Every fact links back to the source it came from.

Vendor: Clawhubvendor · observed Oct 11, 2026
Protocol compatibility: OpenClawcompatibility · observed Oct 11, 2026
Adoption signal: 1K downloadsadoption · observed Oct 11, 2026
Latest release: 2.0.1release · observed May 25, 2026
Handshake status: UNKNOWNsecurity

Install and run

Setup complexity: low.

clawhub skill install s17ezpw1yngek3v1p8tw828bex84rra9:subagent-orchestrator-whhh

Setup complexity is classified as HIGH. You must provision dedicated cloud infrastructure or an isolated VM. Do not run this directly on your local workstation.
Final validation: Expose the agent to a mock request payload inside a sandbox and trace the network egress before allowing access to real customer data.

Contract: missing

curl -s "https://www.xpersona.co/api/v1/agents/clawhub-whhh1994-subagent-orchestrator-whhh/snapshot"

Run-check

$0.02 USD

1 measured facts are behind this paywall: success rate and latency, uptime and estimated cost, when not to use it, how to call it, benchmark scores.

Agents pay $0.02 in USDC. A card payment is $0.50, the smallest a card allows.

Documentation

CLAWHUB

77,788 characters of source documentation, loaded on request.

Extracted files

4 files captured from the source.

SKILL.md

---
name: subagent-orchestrator
description: 将复杂大型任务拆解为多个子任务，通过 Subagent 并行/串行执行，三文件持久化防丢失，独立任务空间管理，Plan 驱动全程，交付包含中间产物与 diff。降低触发阈值——任务步骤超过3步、涉及2个以上来源或平台、context有膨胀迹象时即触发，不等"大型"才拆。触发词：任务分工、子任务编排、多agent协作、上下文溢出、任务分流、context快满了。
metadata: { "openclaw": { "emoji": "🔀", "requires": ["sessions_spawn", "sessions_send", "sessions_list", "write", "read", "exec", "cron", "sessions_yield", "subagents"] }, "version": "2.0.0", "updatedAt": "2026-05-25" }
---

# Subagent Orchestrator Skill

## 版本

- **v2.0.0 (2026-05-25):** 重大升级（参考 Trae Solo / Windsurf Cascade / Devin）
  1. 新增 **Phase 0.5 需求完善**——接需求后主动询问 3-5 个问题，明确后再拆解
  2. 新增 **独立任务空间**——每个任务专属目录，中间文件和最终产物集中管理
  3. 新增 **Plan 驱动机制**——Phase 1 输出完整 Plan，Phase 3 按 Plan 逐项打勾
  4. 改造 **交付清单**——最终交付含中间产物列表 + 相关文件的 diff
  5. 新增 **Checkpoint + Revert**——每步完成后打 checkpoint，支持回退
  6. 新增 **Queued Messages**——执行期间可追加指令，进入队列依次执行
  7. 新增 **实时进度反馈**——每步完成后推送微信，不等最终完成

---

## 核心原则

### 1. 分工原则

| 角色 | 职责 | 不做什么 |
|------|------|----------|
| **Main Agent** | 需求确认、三文件准备、启动 Subagent、监控进度、失败汇总、Queued Messages 路由、最终整合推送微信 | 不亲自执行信息收集/重IO操作 |
| **Subagent** | 执行具体子任务、写数据文件、更新状态、Checkpoint、失败5次即停+报告 | 不做最终决策、不回写任务文件 |

串行：共享资源（浏览器/数据库）；并行：独立资源（web_fetch/不同API）。不确定时默认串行。

### 2. 三文件分离原则

**Token 节流规范：**
- `_任务.md`：Main Agent 写一次，Subagent 启动时只读一次，永不重读
- `_状态.md`：Subagent 增量写，Main Agent 只扫进度，保持 <500字
- `_数据.md`：Subagent 增量写，Main Agent **仅读此文件汇总**
- 失败先写状态文件；同一错误连续失败5次立即停止，不空转

### 3. 独立任务空间原则

每个任务有专属目录，所有产物集中管理：

```
~/.openclaw/workspace/tasks/[任务ID]/
├── _任务.md           # Main Agent 写入，Subagent 只读
├── _状态.md           # Subagent 增量写
├── _数据.md           # Subagent 增量写
├── _交付清单.md       # Phase 4 由 Main Agent 汇总
├── _Plan.md           # 独立 Plan 文件，Subagent 执行前只读此文件（不读 _任务.md）
├── _Checkpoints.md    # Checkpoint 记录，支持回退
├── _MessageQueue.md   # Queued Messages 队列
└── workspace/         # 任务执行时的所有中间文件
    ├── [子任务A]/
    └── ...
```

- **用户可指定空间：** `--workspace /path/to/dir`，则使用用户指定目录
- **命名规则：** `[任务ID]` = `任务名-日期-序号`，如 `ai-research-20260525`

### 4. Plan 驱动原则

Phase 1 的核心产出是 Plan，所有后续执行必须严格按 Plan 推进：

```
## Plan（共 N 步，控制在 3 轮以内）
- [ ] Step 1: [描述]（工具：[预期工具]，产出：[中间/最终产物]）
- [x] Step 2: [描述] ✅ 完成于 10:02（工具：[实际用到的工具]，产出：[实际产出]）
- [ ] Step 3: [描述]
```

- Plan 总步骤数控制在 **3 轮以内**
- Phase 3 每完成一个 Step → 在 Plan 中打 `[x]` 并标注实际工具和产出
- Subagent 发现 Plan 有缺陷 → 写状态文件 + `tag=need_user` → Main Agent 询问用户是否更新

### 5. 需求完善原则（Phase 0.5）

收到任务后，**不立即拆解**，先主动完善需求，询问 3-5 个关键问题：

```
1. 【目标明确化】最终交付物？格式？位置？
2. 【范围界定】包含什么？明确排除什么？
3. 【约束条件】时间/格式/工具限制？参考文件？
4. 【验收标准】怎么算完成？量化指标？
5. 【优先级】必须 vs 可精简？
6. 【背景/上下文】解决什么问题？历史背景？
7. 【工作空间】指定目录 or 默认 tasks/[任务ID]？
```

**跳过条件：** 用户需求已包含完整信息（目标/范围/验收标准/交付位置均有）→ 直接进入 Phase 1

### 6. Checkpoint + Revert 原则

每次 Phase/Step 完成后自动记录 checkpoint，支持回退：

```markdown
## Checkpoint 记录
| ID | 时间戳 | Phase/Step | 状态摘要 | 可回退 |
|-----|---------|-----------|---------|-------|
| ckpt-0 | 10:00 | Phase 1 | Plan已确认，3步子任务 | 是 |
| ckpt-1 | 10:15 | Phase 3 Step1 | 收集阶段完成，10条记录 | 是 |
```

**回退操作步骤（Main Agent 执行）：**
```
1. 确认 ckpt-ID 存在且「可回退

_meta.json

{
  "ownerId": "kn7fgm829te36r33ybgx1v7jpx82khse",
  "slug": "subagent-orchestrator-whhh",
  "version": "2.0.1",
  "publishedAt": 1779697892293
}

references/paths.md

# 路径规范（subagent-orchestrator）

本文档列出 skill 中所有引用的路径和资源，供验证和快速查阅。

## 任务文件路径

| 文件 | 路径规范 | 用途 |
|------|---------|------|
| 任务文件 | `~/.openclaw/workspace/tasks/[子任务]_[日期]_任务.md` | Main Agent 写一次，Subagent 只读 |
| 状态文件 | `~/.openclaw/workspace/tasks/[子任务]_[日期]_状态.md` | Subagent 增量写，Main Agent 快速扫 |
| 数据文件 | `~/.openclaw/workspace/tasks/[子任务]_[日期]_数据.md` | Subagent 增量写，Main Agent 汇总 |
| 最终产出 | `~/.openclaw/workspace/tasks/[项目]_[日期]_最终.md` | Phase 4 输出，用户指定的最终交付 |

## SessionKey 命名规范

格式：`[项目]-[子任务简称]-[YYYYMMDD]`
示例：`travel-xiaohongshu-20260426`

## 依赖 skill

| Skill | 路径 | 触发时调用 |
|-------|------|-----------|
| xiaohongshu-crawler | `skills/xiaohongshu-crawler/SKILL.md` | 小红书攻略抓取 |
| wechat-article-spider | `skills/wechat-article-spider/SKILL.md` | 公众号文章爬取 |
| session-cleanup | `skills/session-cleanup/SKILL.md` | 每日 session 清理（辅助功能） |
| agent-browser-clawdbot | `skills/agent-browser-clawdbot/SKILL.md` | 浏览器自动化（需串行） |

## 依赖脚本

| 脚本 | 路径 | 用途 |
|------|------|------|
| session 扫描 | `skills/session-cleanup/scripts/scan_sessions.sh` | session 清理扫描 |

## 验证检查点

执行任务前，Main Agent 应验证路径可达性：

```bash
# 验证 tasks 目录存在
ls ~/.openclaw/workspace/tasks/ 2>/dev/null || mkdir -p ~/.openclaw/workspace/tasks/

# 验证依赖 skill 存在（按需）
ls ~/.openclaw/workspace/skills/[skill-name]/SKILL.md 2>/dev/null && echo "✅ [skill] 可用" || echo "⚠️ [skill] 未安装"
```

## wiki 关联

最终产出可选择性输出到 wiki：
- Wiki 路径：`~/llm-wiki/raw/articles/[目的地]综合攻略.md`（需用户确认）

skill-card.md

## Description:

Coordinates complex agent work by decomposing tasks into plans, spawning subagents, tracking progress in task files, and consolidating outputs, checkpoints, and diffs.

This skill is ready for commercial/non-commercial use.

## Publisher:

[whhh1994](https://clawhub.ai/user/whhh1994)

### License/Terms of Use:

MIT-0

## Use Case:

Developers and agent operators use this skill to split larger, multi-step work into managed subagent tasks with persistent plans, status files, checkpoints, and final delivery summaries. It is intended for complex research, coding, configuration, and data-gathering workflows that benefit from staged execution and recovery points.

### Deployment Geography for Use:

Global

## Known Risks and Mitigations:

Risk: Broad orchestration authority can spawn sessions, run commands, and write task artifacts.

Mitigation: Install only in environments where this level of agent delegation is intended, and review generated plans before subagents run.

Risk: Persistent task files may store sensitive task details under ~/.openclaw/workspace/tasks.

Mitigation: Avoid using the skill for sensitive work unless local storage is acceptable, and clean task workspaces after review.

Risk: Progress and final summaries may be sent through configured external channels.

Mitigation: Confirm channel routing before use and disable or restrict message behavior where the host environment allows it.

Risk: Cron-related capability may create recurring behavior when enabled with companion cleanup workflows.

Mitigation: Review any scheduled tasks before enabling them and remove schedules that are no longer needed.

## Reference(s):

- [ClawHub skill page](https://clawhub.ai/whhh1994/skills/subagent-orchestrator-whhh)
- [whhh1994 ClawHub profile](https://clawhub.ai/user/whhh1994)
- [Path reference](artifact/references/paths.md)

## Skill Output:

**Output Type(s):** [text, markdown, code, shell commands, configuration, guidance]

**Output Format:** [Markdown task plans, status updates, file paths, command snippets, and delivery summaries]

**Output Parameters:** [1D]

**Other Properties Related to Output:** [May create persistent task artifacts and send progress or final-result messages through the originating channel.]

## Skill Version(s):

2.0.1 (source: server release metadata; source skill frontmatter reports 2.0.0)

## Ethical Considerations:

Users should evaluate whether this skill is appropriate for their environment, review any generated or modified files before relying on them, and apply their organization's safety, security, and compliance requirements before deployment.

Github ReposUpdated 2d agoRank 70

AionUi

Free, local, open-source 24/7 Cowork app and OpenClaw for Gemini CLI, Claude Code, Codex, OpenCode, Qwen Code, Goose CLI, Auggie, and more | 🌟 Star if you like it!

MCPOPENCLAW

Github ReposUpdated 6mo agoRank 70

activepieces

AI Agents & MCPs & AI Workflow Automation • (~400 MCP servers for AI agents) • AI Automation / AI Agent with MCPs • AI Workflows & AI Agents • MCPs for AI Agents

OPENCLAW

Github ReposUpdated 6mo agoRank 70

cherry-studio

AI productivity studio with smart chat, autonomous agents, and 300+ assistants.