name: b2b-daily-automation description: 定时任务自动化 — 早安简报、日报/周报、定时发布、自动化工作流 when_to_use:

• "配置工作日自动化任务（早报 / 开发信 / 社媒 / 每日总结）"
• "用户提到「定时任务」「自动发邮件」「cron」"
• "不要用于：单次执行的任务"

triggers:

• 每日任务
• 自动化
• 定时任务
• 定时发布
• Cron
• 早安简报
• 工作总结
• 晚间总结
• 周报
• 日报

# ... (see skill_registry.py for full list) category: 自动化 version: 1.2.0 author: Foreign Trade Assistant injection_prompt: | 你是 b2b-daily-automation 技能。用于生成每日简报、定时任务等自动化内容。

语言要求：所有输出必须使用中文。 简报、日报、周报、分析摘要等面向用户的文本全部用中文。仅当生成对外邮件/社媒内容时使用目标客户语言。

════════════════════════════════════════ 数据真实性（最重要） ════════════════════════════════════════ 简报/报表中的所有数据必须通过 web_search 实时获取或从数据库查询，不得编造、不得留空、不得使用占位符：

• 汇率 / 金价 / 铜价 / 铝价 / 原油价：逐项搜索（"USD CNY exchange rate today"、"gold price today" 等），每项至少 2 次不同关键词
• 目标市场新闻：搜索每个目标市场的 top business news today；有产品行业配置时补充行业新闻
• 客户跟进 / 待办：从数据库查询（昨日客户互动、A 级待跟进客户）
• KPI / 询盘数 / 业绩：必须来自数据库或用户提供，无数据时标注「N/A — 数据待补充」

════════════════════════════════════════ 任务类型与结构 ════════════════════════════════════════

• 早安简报：汇率（中行中间价）→ 大宗商品 → 目标市场动态 → 客户跟进提醒 → 今日待办 → 今日一句话
• 晚间总结：今日操作回顾 → 定时任务执行情况 → 客户变动 → 待跟进提醒 → 明日建议
• 周报（周五）：执行摘要 → 关键指标 → 管道状态 → 本周成单 → 挑战与阻碍 → 下周计划
• 定时发布：按平台生成可直接发布的内容（LinkedIn / 社媒）

════════════════════════════════════════ cron 任务设置 ════════════════════════════════════════ 用 cronjob 工具创建任务：指定 schedule（cron 表达式）、skills（如 b2b-linkedin-marketing）、deliver 目标；返回任务 ID + 下次执行时间（换算用户本地时区）+ 修改/暂停/删除方法。

质量标准：内容针对用户行业与产品（不泛泛而谈）；生成即用；每份输出带明确的下一步行动建议。

🌅 早安简报规范（Morning Brief Specification）

【最重要】简报中的所有数据必须通过 web_search 实时获取，不得编造、不得留空、不得使用占位符。

执行步骤

1. 使用 web_search 搜索以下信息（逐项搜索，每项至少 2 次不同关键词）：

• 当日美元/欧元/英镑兑人民币汇率（搜索 "USD CNY exchange rate today"）
• 当日金价、铜价、铝价、原油价（搜索 "gold price today" / "copper LME price" / "crude oil price today"）
• 目标市场新闻（搜索每个目标市场的 top business news today）
• 如果用户配置了产品行业（如 "电力金具"），搜索该行业的今日新闻

1. 从数据库查询：

• 昨日对话记录中的客户互动
• 客户管理中的待跟进客户（A 级优先）

1. 输出格式（数值必须来自搜索结果，不得有空白项）：

# 🌅 早安简报 — {日期}
 
## 💱 汇率（中国银行中间价）
| 货币 | 汇率 | 较昨日 |
|------|------|--------|
| 美元 USD | X.XX | +/--X% |
| 欧元 EUR | X.XX | +/--X% |
| 英镑 GBP | X.XX | +/--X% |
（如果 web_search 未返回最新数据，使用已知的最新值并标注 "截至 {日期}"）
 
## 📊 大宗商品
| 商品 | 价格 | 单位 |
|------|------|------|
| 黄金 | XXXX | USD/oz |
| 铜 (LME) | XXXX | USD/t |
| 铝 (LME) | XXXX | USD/t |
| 原油 (WTI) | XX | USD/桶 |
（同理，无最新数据则标注日期）
 
## 🌍 目标市场动态
每个市场至少 2-3 条具体新闻摘要（50-80 字），包含实际公司名、数据、政策变化。
 
## 👤 客户跟进提醒
从客户管理表中列出昨日互动过的客户 + A 级待跟进客户，
显示客户名称、国家、最后联系时间、建议跟进内容。
 
## ⏰ 今日待办
基于客户跟进提醒 + 昨日对话生成的 3-5 条具体可执行事项。
 
## 💪 今日一句话

Generated by Foreign Trade Assistant | Tasks for your review

 
### Task 2: LinkedIn Content (09:30 - Workdays)
 
**Purpose**: AI generates content ready to copy-paste to LinkedIn
 
**Content Types** (rotating):
- Text post with insight
- Document post (guide/checklist)
- Poll question
- Industry observation
 
**Generated Content Template**:

LinkedIn Post — Ready to Publish

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

[Content — optimized for LinkedIn engagement]

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

📝 Caption: [Full post text with hashtags]

🏷️ Hashtags:

#[Industry] #[Product] #[B2B] #[Sourcing] #[CompanyName]

⏰ Best time to post: 09:30 AM local time ✅ Ready to paste and publish

 
### Task 3: European Outreach (10:00 - Workdays)
 
**Purpose**: Send targeted emails to European prospects/customers
 
**Target Selection Logic**:
1. Filter customers in European markets
2. Check last contact date
3. Select based on follow-up sequence
4. Personalize each email
 
**Generated Email Template**:

TO: [Recipient Email] SUBJECT: [Personalized subject line]

Hi [Name],

[Personalized opening — reference something specific about them]

[Body — value proposition, specific to their industry/needs]

[Call to action — specific, easy to respond to]

Best regards, [Your Name] [Title] | [Company] [Phone] | [WeChat]

 
### Task 4: B2B Platform Check (11:00 - Workdays)
 
**Purpose**: Review and respond to platform inquiries
 
**Check List**:

ALIBABA INTERNATIONAL STATION: □ New RFQ/ inquiries: [X] □ Pending quotes to follow up: [X] □ Product views/lead changes: [X] □ Competitor activity: [Notes]

MADE-IN-CHINA / OTHER PLATFORMS: □ New messages: [X] □ Action items: [Notes]

ACTION ITEMS FOR TODAY:

1. [ ] Respond to [X] new inquiries
2. [ ] Follow up on [X] pending quotes
3. [ ] [Other task]

 
### Task 5: Social Media Posting (14:00 - Workdays)
 
**Purpose**: Generate and schedule social content
 
**Platform Rotation**:
| Day | Platform | Content Type |
|-----|----------|-------------|
| Mon | Facebook | Industry article/insight |
| Mon | Instagram | Product showcase |
| Tue | TikTok | Factory/Behind-the-scenes |
| Tue | Facebook | Poll or question |
| Wed | Instagram | Reel (product demo) |
| Wed | YouTube | (Weekly) Long video |
| Thu | TikTok | Trending sound adaptation |
| Thu | Facebook | Behind-the-scenes |
| Fri | Instagram | Stories (recap) |
| Fri | TikTok | Quick tip |
 
**Generated Content Template**:

Social Media Post — Ready to Publish

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

Platform: [Facebook/Instagram/TikTok/YouTube] Scheduled time: [14:00 local time]

📝 Caption/Description: [Full post text]

🏷️ Hashtags: [Platform-specific hashtags]

🖼️ Media: [Description of image/video to use]

✅ Ready to schedule or publish

 
### Task 6: US/North America Outreach (16:00 - Workdays)
 
**Purpose**: Send targeted emails to Americas prospects/customers
 
**Same structure as European Outreach (Task 3), but:
- Target: North America, Latin America
- Timing: Catch end of their workday
 
### Task 7: Daily Summary (17:30 - Workdays)
 
**Purpose**: End-of-day recap driven by real data — chat history, cron outputs, and customer records.
 
**Data Sources** (use tools to query, never fabricate):
 
1. **今日聊天记录** — 使用 `database` 工具查询 `conversations` 表：
   `SELECT query, response, created_at FROM conversations WHERE company_id = {当前公司ID} AND date(created_at) = date('now', 'localtime') ORDER BY id ASC`
   从这些对话中提取今天做了什么：客户背调、开发信、文档生成、平台诊断等。
 
2. **今日定时任务输出** — 使用 `list_files` 或 `read_file` 读取 `~/.hermes/cron/output/{today}/` 目录下的任务输出文件，汇总每个定时任务产出了什么。
 
3. **客户变动** — 使用 `database` 工具查询今日新增/更新的客户：
   `SELECT name, created_at FROM customers WHERE company_id = {当前公司ID} AND date(created_at) = date('now', 'localtime')`
 
4. **待跟进提醒** — 从今日聊天记录中找出涉及具体客户名称的对话，提取需要明天跟进的事项。
 
**Output Rules**:
- **所有数据必须来自工具查询结果**，没有数据则写"今日无记录"
- **禁止编造**任何数字、客户名、完成事项
- 简明扼要，每条 1-2 行，不需要长篇描述
 
**Generated Content Template**:

📊 每日工作总结 — {日期}

今日操作回顾

（从今日聊天记录中提取实际完成的操作类型和数量）

• 客户背调: X 项
• 开发信/邮件: X 封
• 文档生成: X 份
• 平台诊断: X 次
• 其他: ...

定时任务执行情况

（从 cron output 目录读取今日各任务的输出，简述每个任务产出）

• 早安简报: 已生成 / 未执行
• LinkedIn 加人: 已生成 X 个搜索方案
• ...

今日客户变动

• 新增客户: （列出名字，无则写"无"）
• 更新客户: （列出名字和更新内容摘要）

待跟进提醒

（从聊天记录中提取涉及客户沟通的内容）

明日建议

（基于今天未完成的事项和待跟进客户，给出 3-5 条具体建议）

Generated by Foreign Trade Assistant

 
## Phase 3: Weekly Report (Friday 17:30)
 
**Purpose**: Weekly business review and planning
 
**Generated Content Template**:

📈 Weekly Report — Week [X] ([Start Date] - [End Date])

Executive Summary

[2-3 sentences: How was this week overall?]

Key Metrics

Pipeline Status

Top Wins This Week

1. [Win 1]: [Description — customer, product, value]
2. [Win 2]: [Description]

Challenges & Blockers

Customer Spotlight

🏆 Top Customer: [Name]

• This week: [Activity]
• YTD revenue: [Amount]

⚠️ At-Risk: [Name]

• Risk: [Description]
• Action taken: [Action]

Next Week's Plan

Priorities

1. [ ] [Priority 1]
2. [ ] [Priority 2]
3. [ ] [Priority 3]

Key Dates

• [Meeting/Call on Day]
• [Quote follow-up due]
• [Sample delivery expected]

Social Media Performance

Week [X] of [Year] | Prepared by Foreign Trade Assistant

 
## Phase 4: Cron Job Setup
 
### Task Scheduling Template
 
Use Hermes cron to schedule:
 

Morning Brief — 08:00 Monday-Friday

trade cron create \ --name "Morning Brief" \ --schedule "0 8 1-5" \ --deliver local \ --prompt "按照 b2b-daily-automation 技能中的「早安简报规范」生成今日简报。必须使用 web_search 逐项搜索实时汇率（USD/EUR/GBP→CNY）、大宗商品价格（金/铜/铝/原油）、目标市场新闻。查询昨日对话记录和客户待跟进清单。所有数值必须来自搜索结果，不得编造或留空。输出格式严格遵守简报规范模板。"

LinkedIn Post — 09:30 Monday-Friday

trade cron create \ --name "LinkedIn Post" \ --schedule "30 9 1-5" \ --deliver local \ --prompt "Generate LinkedIn post ready to publish. Include: engaging text post with hook, value proposition, and CTA. Add 5 relevant hashtags. User is in [INDUSTRY] exporting [PRODUCT]. Content should be professional but authentic."

European Outreach — 10:00 Monday-Friday

trade cron create \ --name "EU Outreach" \ --schedule "0 10 1-5" \ --deliver local \ --prompt "Generate [X] personalized outreach emails for European prospects/customers in [COUNTRIES]. For each: reference something specific about them, explain value proposition, include clear CTA. Check CRM for follow-up sequence."

Platform Check — 11:00 Monday-Friday

trade cron create \ --name "Platform Check" \ --schedule "0 11 1-5" \ --deliver local \ --prompt "Generate platform review checklist for today. Include: new RFQ count, pending quotes requiring follow-up, product view trends, competitor activity notes. Platforms: [ALIBABA/INDIAMART/ETC]."

Social Media — 14:00 Monday-Friday

trade cron create \ --name "Social Media Post" \ --schedule "0 14 1-5" \ --deliver local \ --prompt "Generate social media post for [PLATFORM]. Today's content type: [TEXT/IMAGE/VIDEO/POLL]. Include: caption, hashtags, media suggestions. User exports [PRODUCT] targeting [AUDIENCE]."

US Outreach — 16:00 Monday-Friday

trade cron create \ --name "US Outreach" \ --schedule "0 16 1-5" \ --deliver local \ --prompt "Generate [X] personalized outreach emails for US/Americas prospects/customers. Same structure as EU outreach. Target markets: [US/CANADA/LATAM]."

Daily Summary — 17:30 Monday-Friday

trade cron create \ --name "Daily Summary" \ --schedule "30 17 1-5" \ --deliver local \ --prompt "生成每日工作总结。按以下步骤从实际数据中收集信息，不编造任何内容：1) database 工具查询 conversations 表今天的所有对话，提取实际操作类型；2) read_file 读取 ~/.hermes/cron/output/ 下今天各任务的输出文件；3) database 工具查询 customers 表今天新增/更新的客户；4) 从聊天记录中提取涉及客户沟通的内容生成待跟进提醒。所有数据必须来自工具查询结果，没有数据则写'今日无记录'。遵循 b2b-daily-automation SKILL.md 中 Task 7 的模板格式。"

 
### Cron Schedule Reference
 
| Schedule | Meaning |
|----------|---------|
| `0 8 * * 1-5` | 8:00 AM, Monday-Friday |
| `30 9 * * 1-5` | 9:30 AM, Monday-Friday |
| `0 10 * * 1-5` | 10:00 AM, Monday-Friday |
| `0 14 * * 1-5` | 2:00 PM, Monday-Friday |
| `0 16 * * 1-5` | 4:00 PM, Monday-Friday |
| `30 17 * * 1-5` | 5:30 PM, Monday-Friday |
| `0 18 * * 5` | 6:00 PM, Friday (Weekly report) |
 
## Phase 5: Integration with CRM
 
### Data Flow
 

┌─────────────┐ ┌─────────────┐ ┌─────────────┐ │ Cron │───▶│ Trade │───▶│ User │ │ Tasks │ │ Assistant │ │ Executes │ └─────────────┘ └─────────────┘ └─────────────┘ │ │ │ │ ▼ │ │ ┌─────────────┐ │ │ │ CRM │◀──────────┘ │ │ Updates │ └──────────▶│ (next day)│ └─────────────┘

 
### What Gets Logged
 
After each task execution:
- Emails sent → logged to CRM
- Follow-ups completed → logged to CRM
- Social posts published → logged to analytics
- Inquiries received → added to pipeline
 
## Phase 6: Task Customization
 
### Customizing for Your Business
 
**Adjust timing based on**:
- Your working hours
- Target market time zones
- Personal energy patterns (do creative work when you're fresh)
 
**Adjust content based on**:
- Industry (manufacturing needs more technical content)
- Product type (complex products need more education)
- Target audience (engineers vs. buyers vs. C-suite)
 
**Adjust frequency based on**:
- Platform activity levels
- Customer engagement rates
- Your capacity to execute
 
## Quality Standards
 
1. **内容真实性**：KPI 数据、询盘数量、业绩统计必须来自数据库查询或用户提供，不编造任何指标值。无数据时标注 "N/A — 数据待补充"
2. **Relevance**: Every piece of content must be specific to the user's industry and products
2. **Timeliness**: Posts go out during optimal engagement windows for target audiences
3. **Personalization**: Outbound messages reference specific prospect details
4. **Completeness**: Generated content is ready to use — no editing needed
5. **Consistency**: Same quality and tone across all automated outputs
6. **Actionability**: Every brief and summary has clear next steps
 
## Common Pitfalls
 
1. **Setting and forgetting**: Review automated outputs weekly, adjust based on results
2. **Generic content**: Without specific product info, content will be generic — provide details
3. **Too much automation**: Balance automated content with genuine personal engagement
4. **Ignoring analytics**: Track what's working and optimize timing/content
5. **Losing personal touch**: Automation handles routine; personal relationships need human contact
6. **Not following up on generated leads**: If you generate an inquiry response but don't follow up, automation is wasted