Ai Company

完全自主的AI公司运营系统 - 7×24小时自动化发现需求、设计、开发、销售、运维，实现盈利的轻量级解决方案

Security Scan

The skill's documentation and example code clearly expect multiple API keys, network access, and automated posting/scheduling, but the registry metadata does not declare those requirements — this mismatch and the broad runtime actions warrant caution before installing or running.

Purpose & Capability

The skill claims to be a lightweight, file-based AI company framework, which is plausible, but the SKILL.md and examples expect many external integrations (Anthropic/Claude, GitHub, Twitter, SMTP) and instruct creating cron jobs and automating posts/operations. The registry metadata declares no required env vars, credentials, or binaries despite the project repeatedly referencing API keys and tokens (ANTHROPIC_API_KEY, GITHUB_TOKEN, TWITTER_*, SMTP_*). That mismatch is incoherent: a skill that scans and posts to external platforms legitimately needs API credentials and network access and should declare them.

Instruction Scope

Runtime instructions tell an agent/user to scan GitHub/Reddit/Twitter, create repositories, send marketing emails, and run recurring cron tasks. The SKILL.md and examples explicitly reference reading/writing .env and config.yaml and writing shared data and logs. Those instructions go beyond purely local offline tasks — they direct actions that can access external services and perform outbound communication (social posting, emailing, GitHub operations). The instructions also give the agent broad discretion to discover and act on 'opportunities', which could lead to automated actions impacting external accounts.

Install Mechanism

There is no formal install spec in the registry (instruction-only), which minimizes supply-chain risk from downloads. However SKILL.md and examples require installing third‑party Python packages (anthropic, requests, python-dotenv, pyyaml). Those are standard but the project will rely on networked SDKs; users should vet and pin versions before pip installing in production.

Credentials

The package examples and config files expect multiple sensitive environment variables (Anthropic API key, GitHub token, SMTP credentials, Twitter keys) and suggest storing them in .env — yet the registry lists none. Requesting many high‑value secrets is proportionate only if the user intends the described integrations; the omission from metadata means the skill could be installed without the user being warned about these credential needs. That increases the chance someone will run the code with overly powerful or improperly scoped tokens.

Persistence & Privilege

always:false and no special OS restrictions are appropriate. The instructions recommend setting cron jobs and starting a long-running 'main.py' scheduler which gives the project persistent, periodic execution on a host. This is normal for automation but does create ongoing capability to act (post, email, scan). The skill does not claim to modify other skills or system-wide settings, but setting up cron jobs and automated network activity increases operational impact and should be considered before enabling.

What to consider before installing

Key things to consider before installing or running this skill: - Metadata mismatch: The registry declares no required credentials, but the SKILL.md and example config expect many sensitive API keys (Anthropic/Claude, GitHub, Twitter, SMTP). Treat that as a red flag — you should only run this if you understand and accept providing those tokens. - Least privilege for tokens: If you try it, create tokens with the minimal scopes needed (e.g., read-only GitHub scopes when scanning, limited Twitter/posting account, SMTP account dedicated to this project). Do not use admin or organization-level tokens. - Review code before running: The included examples are simple and mostly local, but they do instruct networked operations and scheduling. Inspect the examples (init script, main.py, employee scripts) and ensure there are no endpoints or hidden uploads you don't expect. - Run in a sandbox first: Test in an isolated environment (container or VM) with mock APIs or with 'mock_apis' / development toggles enabled. Avoid running against production accounts until you're confident. - Protect secrets and rotate keys: Keep .env out of version control, rotate credentials after testing, and consider using a secrets manager rather than plaintext .env files. - Consider operational risks: The system can automatically post to social media and send emails. Ensure approval workflows or human-in-the-loop controls are in place to prevent unwanted public communications. If you want to proceed safely, ask the skill author to update registry metadata to declare required env vars and to provide minimal example config that uses mock integrations by default.

AI Company 自动化运营技能

概述

重要说明：这是一个技能定义，不是完整的项目实现。使用本技能来创建和运行您的AI公司。

这个技能教你如何构建一个完全由AI员工组成的公司，实现：

• 自主发现需求：扫描GitHub、Reddit、Twitter等平台发现机会
• 智能设计开发：AI产品设计师和开发者团队协作
• 自动化销售：AI销售和营销自动获取客户
• 持续交付支持：AI客服和DevOps自动运维
• 数据驱动优化：基于反馈持续迭代产品和流程
• 版本化管理：所有AI员工可版本控制和快速回滚

核心特点

1. 去中心化AI员工网络

每个AI员工都是独立的智能体，通过事件总线协作，无单点故障：

机会发现层 → 产品设计层 → 开发交付层 → 商业运营层 → 监控优化层

2. 持续优化循环

系统不断学习和改进：

发现机会 → 开发产品 → 获取客户 → 收集反馈 → 分析学习 → 产品迭代 → 重复

3. 轻量级技术栈

只需Python + JSON文件，无需复杂的基础设施：

- Python 3.10+
- Claude Agent SDK
- 简单的JSON文件存储
- cron定时任务
- 可选GitHub Actions

4. 人类监督保障

AI监控AI，异常时自动告警人类：

自我监控 → 同伴监控 → 人类监控面板 → 介入决策

AI员工角色

Market Research AI（市场研究专家）

职责：

• 扫描GitHub Issues发现技术痛点
• 分析Reddit和Hacker News讨论
• 监控Twitter技术趋势
• 追踪竞品动向
• 评估市场机会和收入潜力

输出：opportunities.json - 包含市场机会、痛点分析、潜在收入

Product Designer AI（产品设计师）

职责：

• 将机会转化为产品概念
• 设计MVP功能集
• 制定定价策略
• 创建产品路线图

输出：product_designs.json - 产品设计文档、功能列表、定价模型

Developer AI（开发专家）

职责：

• 实现产品功能
• 编写技术文档
• 创建自动化测试
• 修复bug和性能优化
• 管理代码仓库

输出：GitHub仓库、文档、测试套件

Sales & Marketing AI（销售营销专家）

职责：

• 生成营销内容
• 管理社交媒体账号
• 回复客户咨询
• 跟进销售线索
• 维护客户关系

输出：营销活动、销售记录、客户数据库

Customer Support AI（客服专家）

职责：

• 回答客户问题
• 解决技术问题
• 收集产品反馈
• 识别常见问题并改进FAQ

输出：支持工单、客户反馈、知识库更新

Monitor AI（监控优化专家）

职责：

• 监控所有AI员工状态
• 检测性能异常
• 生成优化建议
• 触发人类告警

输出：健康报告、告警、优化建议

Finance AI（财务专家）

职责：

• 追踪收入和支出
• 计算利润率
• 生成财务报告
• 建议定价调整

输出：财务报告、收入分析、趋势预测

技能结构 vs 项目结构

技能文件结构（当前）

ai-company/                 # 技能定义目录
├── SKILL.md                # 技能主文档
├── README.md               # 项目说明
├── LICENSE                 # 许可证
├── CONTRIBUTING.md         # 贡献指南
├── docs/                   # 详细文档
│   ├── design.md          # 设计文档
│   └── api.md             # API文档
└── examples/              # 示例代码
    ├── simple_ai_employee.py
    ├── simple_event_bus.py
    ├── simple_coordinator.py
    └── config.yaml

使用本技能创建的项目结构

my-ai-company/              # 使用技能创建的项目
├── employees/              # AI员工实现
│   ├── market_researcher.py
│   ├── product_designer.py
│   ├── developer.py
│   ├── sales_marketing.py
│   ├── customer_support.py
│   ├── monitor.py
│   └── finance.py
├── prompts/                # AI员工提示词（版本化）
│   ├── market_researcher/
│   │   ├── v1.0.md
│   │   └── v1.1.md
│   ├── sales_marketing/
│   │   ├── v1.0.md
│   │   ├── v2.0.md
│   │   └── v2.1.md
│   └── versions.json
├── shared/                 # 共享数据
│   ├── opportunities.json
│   ├── products.json
│   ├── customers.json
│   ├── sales.json
│   ├── state.json
│   └── metrics.json
├── workflows/              # 工作流定义
│   ├── discover_opportunities.yaml
│   ├── build_product.yaml
│   ├── make_sale.yaml
│   └── optimize_system.yaml
├── logs/                   # 日志文件
├── main.py                 # 主调度器
└── config.yaml             # 配置文件

快速开始

1. 安装依赖

pip install anthropic python-dotenv pyyaml requests

2. 创建AI公司项目

# 方法1：使用初始化脚本（推荐）
cd skills/ai-company/examples
python3 init_ai_company.py my-ai-company

# 方法2：手动创建
mkdir my-ai-company
cd my-ai-company
# 按照项目结构手动创建目录和文件

3. 配置API密钥

cd my-ai-company
cp .env.example .env
# 编辑.env，添加你的API密钥
nano .env

4. 运行示例测试

# 测试AI员工示例
python3 ../examples/simple_ai_employee.py

# 测试完整工作流示例
python3 ../examples/simple_coordinator.py

5. 启动你的AI公司

# 启动AI团队
python main.py start

# 查看状态
python main.py status

# 停止AI团队
python main.py stop

5. 设置定时任务

crontab -e
# 添加：
*/30 * * * * cd /path/to/my-ai-company && python main.py --task discover_opportunities
0 9 * * * cd /path/to/my-ai-company && python main.py --task daily_optimization
*/15 * * * * cd /path/to/my-ai-company && python main.py --task health_check

配置文件

config.yaml

company:
  name: "My AI Company"
  industry: "software_development"
  target_market: "individuals_small_business"

ai_employees:
  - name: market_researcher
    enabled: true
    version: "v1.1"
  - name: product_designer
    enabled: true
    version: "v1.0"
  - name: developer
    enabled: true
    version: "v1.0"
  - name: sales_marketing
    enabled: true
    version: "v2.1"
  - name: customer_support
    enabled: true
    version: "v1.0"
  - name: monitor
    enabled: true
    version: "v1.0"
  - name: finance
    enabled: true
    version: "v1.0"

apis:
  anthropic_api_key: "${ANTHROPIC_API_KEY}"
  github_token: "${GITHUB_TOKEN}"

schedule:
  opportunity_discovery: "*/30 * * * *"
  daily_optimization: "0 9 * * *"
  health_check: "*/15 * * * *"

monitoring:
  alert_email: "[e-mail]"
  alert_threshold:
    error_rate: 0.1
    revenue_drop: 0.2

工作流程

1. 机会发现流程

触发条件：每30分钟
流程：
  1. Market Research AI扫描多个平台
  2. 分析和评分每个机会
  3. 保存高价值机会到opportunities.json
  4. 发布opportunity.discovered事件

2. 产品开发流程

触发条件：新机会发现
流程：
  1. Product Designer AI设计产品
  2. Developer AI实现MVP
  3. QA自动测试
  4. 部署到生产环境
  5. 发布product.ready事件

3. 销售流程

触发条件：产品就绪
流程：
  1. Marketing AI创建营销内容
  2. 多渠道推广（Twitter、Reddit、邮件）
  3. Sales AI回复咨询
  4. 跟进线索
  5. 成交记录

4. 优化流程

触发条件：每天早上9点
流程：
  1. 分析昨天的数据
  2. 识别问题和机会
  3. 优先级排序
  4. 执行改进：
     - 产品迭代
     - 营销优化
     - 定价调整
     - 客户挽回
  5. 学习和记录

版本控制和A/B测试

AI员工版本管理

# 创建新版本
python main.py --new-version sales_agent v2.2

# A/B测试
python main.py --ab-test sales_agent v2.1 v2.2 --traffic 0.2

# 查看测试结果
python main.py --ab-test-results

# 回滚
python main.py --rollback sales_agent

提示词版本化

所有AI员工的提示词都纳入版本控制：

prompts/sales_marketing/
├── v1.0.md     # 初始版本
├── v2.0.md     # 重大更新
└── v2.1.md     # 当前版本

监控和告警

实时监控

# 查看所有AI员工状态
python main.py --status

# 查看特定AI的日志
tail -f logs/market_researcher.log

# 启动Web仪表板
python main.py --dashboard
# 访问 http://localhost:5000

告警级别

• INFO: 正常运行
• WARNING: 性能下降，需关注
• ERROR: 任务失败，自动重试中
• CRITICAL: 需要人类介入

告警触发条件

• 同一AI连续失败3次
• 收入下降超过20%
• 客户投诉率上升
• 系统资源使用超过90%
• AI检测到无法处理的异常

数据存储

JSON数据结构

opportunities.json

{
  "opportunities": [
    {
      "id": "opp_001",
      "source": "reddit/r/webdev",
      "pain_point": "缺少自动化测试工具",
      "potential_revenue": 500,
      "difficulty": "medium",
      "market_size": "large",
      "status": "validated"
    }
  ]
}

products.json

{
  "products": [
    {
      "id": "prod_001",
      "name": "AutoTest Pro",
      "version": "2.3.0",
      "status": "active",
      "pricing": {"starter": 29, "pro": 99},
      "metrics": {
        "daily_sales": 15,
        "refund_rate": 0.02,
        "customer_satisfaction": 4.5
      }
    }
  ]
}

customers.json

{
  "customers": [
    {
      "id": "cust_001",
      "name": "John Doe",
      "email": "[e-mail]",
      "status": "active",
      "lifetime_value": 590,
      "health_score": 0.8
    }
  ]
}

最佳实践

1. 从小开始

• 先启动1-2个AI员工
• 验证工作流程
• 逐步扩展到完整的AI团队

2. 人工监督初期

• 前几周密切关注AI决策
• 定期审查AI输出
• 调整提示词和配置

3. 数据驱动优化

• 定期查看指标和报告
• 基于数据做决策
• A/B测试重大变更

4. 版本控制一切

• 所有提示词纳入Git
• 重大变更前打标签
• 保持快速回滚能力

5. 客户体验优先

• 快速响应客户咨询
• 主动收集反馈
• 持续改进产品质量

故障排查

AI员工不工作

# 检查状态
python main.py --status

# 查看日志
tail -f logs/<employee_name>.log

# 重启AI员工
python main.py --restart <employee_name>

性能下降

# 查看优化建议
python main.py --optimizations

# 检查资源使用
python main.py --resources

# 回滚到上一版本
python main.py --rollback <employee_name>

收入异常

# 查看财务报告
python main.py --financial-report

# 分析销售数据
python main.py --analyze-sales

# 检查客户健康度
python main.py --customer-health

高级功能

自定义AI员工

# employees/custom_ai.py
from ai_employee import AIEmployee

class CustomAI(AIEmployee):
    name = "custom_ai"
    role = "自定义专家"

    tools = [
        'custom_tool_1',
        'custom_tool_2'
    ]

    def process(self, task):
        # 自定义处理逻辑
        result = self.claude.process(task, self.tools)
        return result

自定义工作流

# workflows/custom_workflow.yaml
name: 自定义工作流
triggers:
  - cron: "0 */2 * * *"
steps:
  - step_1:
      ai: custom_ai
      action: custom_action
  - step_2:
      ai: another_ai
      action: another_action

集成外部服务

# config.yaml
integrations:
  slack:
    webhook_url: "https://hooks.slack.com/..."
  discord:
    bot_token: "your-bot-token"
  email:
    smtp_server: "smtp.gmail.com"
    smtp_port: 587

扩展阅读

• 设计文档 - 详细的系统设计
• API文档 - 完整的API参考
• 示例项目 - 实际运行的AI公司
• 社区讨论 - 分享经验和技巧

贡献

欢迎贡献！请查看 CONTRIBUTING.md 了解如何参与。

许可证

MIT License - 详见 LICENSE

---

作者: AI CEO Automation Team 版本: 2.0.0 最后更新: 2024-03-09

开始构建你的AI公司吧！ 🚀