department-manager

管理按部门组织的人工智能子代理团队。当您需要将任务委派给专门代理、跟踪部门输出、分配角色以及协调多代理工作流程时使用。对于运行业务或复杂项目的自主代理至关重要。

# Department Manager

Organize your AI workers into departments. Assign tasks, track output, and coordinate multi-agent teams like a CEO.

## Why This Exists

Running an autonomous business or complex project means juggling multiple tasks across different domains: content, research, code, marketing, operations. Instead of doing everything sequentially, organize your workers into departments and delegate in parallel.

## Commands

### Create a department
```bash
python3 {baseDir}/scripts/departments.py create --name "content" --description "SEO blog posts, marketing copy, newsletters" --model "arcee-ai/trinity-large-preview:free"
```

### List departments
```bash
python3 {baseDir}/scripts/departments.py list
```

### Assign a task to a department
```bash
python3 {baseDir}/scripts/departments.py assign --dept "content" --task "Write a blog post about OpenClaw memory system" --priority high
```

### Check department status
```bash
python3 {baseDir}/scripts/departments.py status --dept "content"
```

### View all active tasks across departments
```bash
python3 {baseDir}/scripts/departments.py active
```

### Complete a task
```bash
python3 {baseDir}/scripts/departments.py complete --task-id 1 --output "Blog post written and saved to drafts/memory-post.md"
```

### Department report (summary of all departments)
```bash
python3 {baseDir}/scripts/departments.py report
```

### Remove a department
```bash
python3 {baseDir}/scripts/departments.py remove --name "content"
```

## Recommended Department Structure

| Department | Model | Responsibilities |
|------------|-------|-----------------|
| content | arcee-ai/trinity-large-preview:free | Blog posts, marketing copy, newsletters |
| research | stepfun/step-3.5-flash:free | Market research, competitor analysis, data gathering |
| engineering | openrouter/free | Code generation, bug fixes, tooling |
| operations | (manual/CEO) | Budget, strategy, communications |
| security | (manual/CEO) | Skill scanning, threat assessment, audits |

## Data Storage

Department data stored in `~/.openclaw/department-manager/departments.json`.

## Tips

- Create departments based on your actual workflow, not hypothetical needs
- Assign one default model per department for consistency
- Review completed tasks before publishing — QA is the CEO's job
- Use priority levels to focus workers on what matters now
- Run `report` at the end of each day for a team overview

---

## 中文说明

# 部门管理器

将你的 AI 工作者组织成多个部门。像 CEO 一样分配任务、跟踪产出并协调多代理团队。

## 为何需要它

运营一项自主业务或复杂项目意味着要同时处理跨不同领域的多项任务：内容、研究、代码、营销、运营。与其按顺序完成所有工作，不如将工作者组织成部门并行委派。

## 命令

### 创建部门
```bash
python3 {baseDir}/scripts/departments.py create --name "content" --description "SEO blog posts, marketing copy, newsletters" --model "arcee-ai/trinity-large-preview:free"
```

### 列出部门
```bash
python3 {baseDir}/scripts/departments.py list
```

### 向部门分配任务
```bash
python3 {baseDir}/scripts/departments.py assign --dept "content" --task "Write a blog post about OpenClaw memory system" --priority high
```

### 检查部门状态
```bash
python3 {baseDir}/scripts/departments.py status --dept "content"
```

### 查看跨部门的所有活动任务
```bash
python3 {baseDir}/scripts/departments.py active
```

### 完成任务
```bash
python3 {baseDir}/scripts/departments.py complete --task-id 1 --output "Blog post written and saved to drafts/memory-post.md"
```

### 部门报告（所有部门的汇总）
```bash
python3 {baseDir}/scripts/departments.py report
```

### 移除部门
```bash
python3 {baseDir}/scripts/departments.py remove --name "content"
```

## 推荐的部门结构

| 部门 | 模型 | 职责 |
|------------|-------|-----------------|
| content | arcee-ai/trinity-large-preview:free | 博客文章、营销文案、电子简报 |
| research | stepfun/step-3.5-flash:free | 市场研究、竞品分析、数据收集 |
| engineering | openrouter/free | 代码生成、缺陷修复、工具链 |
| operations | （手动/CEO） | 预算、战略、沟通 |
| security | （手动/CEO） | 技能扫描、威胁评估、审计 |

## 数据存储

部门数据存储在 `~/.openclaw/department-manager/departments.json` 中。

## 提示

- 根据你的实际工作流程而非假想需求来创建部门
- 为每个部门分配一个默认模型以保持一致性
- 在发布前审查已完成的任务——质量把控是 CEO 的职责
- 使用优先级别让工作者聚焦于当前最重要的事情
- 每天结束时运行 `report` 以获取团队概览