lark-workflow-chat-digest

Group Message Daily Report Generator Workflow

CRITICAL — Before starting, you MUST first use the Read tool to read ../lark-shared/SKILL.md, which contains authentication and permission handling

Applicable Scenarios

• "Help me summarize what was discussed in the project group chat today" / "group daily report"
• "Generate a weekly report for XX group" / "group message summary"
• "What important messages are in this group today" / "group chat summary"
• "Help me check if there are any important notices in the group today" / "today's group message review"
• "Generate a discussion summary for the product group this week"

Prerequisites

Only supports user identity. Ensure authorization before execution:

lark-cli auth login --domain im
# 如需保存文档，还需：
lark-cli auth login --domain im,docs

Workflow

{群名称/群ID + 时间范围}
  │
  ├── im +chat-search ──────────────► 查找群聊（获取 chat_id）
  │
  ├── im +chat-messages-list ───────► 拉取群消息（分页，max 50/页）
  │     └── --start / --end 过滤时间范围
  │
  ├── AI 分析消息 ──────────────────► 提取关键信息
  │     ├── 话题分类
  │     ├── 决策记录
  │     ├── 待办事项
  │     ├── 问题与风险
  │     └── 重要通知
  │
  ├── AI 生成结构化日报 ────────────► Markdown 格式
  │
  ├── [可选] docs +create ──────────► 保存为飞书文档
  │
  └── [可选] im +messages-send ─────► 发送到群聊

---

Step 1: Find the group chat

If the user provided a group name, first get the chat_id:

lark-cli im +chat-search --query "<群名称>" --as user

Extract from the returned result chat_id（oc_xxx format). If the user directly provided a chat_id, skip this step.

Step 2: Fetch group messages

# 拉取指定时间范围内的消息（按时间倒序，最新在前）
lark-cli im +chat-messages-list \
  --chat-id "<chat_id>" \
  --start "2026-04-09T00:00:00+08:00" \
  --end "2026-04-09T23:59:59+08:00" \
  --page-size "50" \
  --sort "asc" \
  --as user

Pagination handling

Each page can contain at most 50 messages. If the returned result contains has_more: true, use --page-token to continue fetching:

# 第一页
lark-cli im +chat-messages-list --chat-id "oc_xxx" --start "..." --end "..." --page-size "50" --as user
# 返回 page_token 后，拉取下一页
lark-cli im +chat-messages-list --chat-id "oc_xxx" --start "..." --end "..." --page-size "50" --page-token "<token>" --as user

Note：

• If the message volume for a day exceeds 500, it is recommended to analyze only the most recent 200-300 messages to avoid exceeding the context
• Filter out pure emoji replies, system messages, and image/file messages (unless the user specifically requests them)
• Use --sort asc to sort in chronological order, making it easier for the AI to understand the discussion flow

Step 3: AI generates the daily report

Hand the fetched messages to the AI for analysis, and generate the daily report according to the following structure:

## {群名} 日报 — {日期}

### 概览
- 消息总数：N 条
- 活跃成员：N 人
- 主要话题：{2-3个关键词}

### 重要决策
| 决策内容 | 提出者 | 时间 |
|---------|--------|------|

### 待办事项
| 事项 | 负责人 | 来源 |
|------|--------|------|

### 话题讨论
#### 1. {话题标题}
- **摘要**：一句话概括
- **参与人**：A、B、C
- **结论**：{如有}

#### 2. {话题标题}
...

### 问题与风险
| 问题描述 | 提出者 | 状态 |
|---------|--------|------|

### 其他通知
- {通知1}
- {通知2}

Analysis rules:

• Ignore meaningless chit-chat and pure emoji messages
• Focus on messages containing decisions, to-dos, and issues
• Preserve key context, but remove redundant discussion
• If the message contains @某人 reminder, mark it in the to-dos

Step 4: Output the daily report (optional)

Send to the group chat

lark-cli im +messages-send \
  --chat-id "<chat_id>" \
  --markdown "{日报内容}" \
  --as bot

Send to an individual

lark-cli im +messages-send \
  --user-id "<open_id>" \
  --markdown "{日报内容}" \
  --as bot

Save as a document

lark-cli docs +create \
  --title "{群名} 日报 — {日期}" \
  --markdown "{日报内容}" \
  --as user

Fallback Strategy

Scenario	Fallback Plan
Group chat not found	Prompt the user to confirm the group name, or use chat_id directly
No messages in the group	Return "No messages in the group during this period"
Message volume too large	Analyze only the most recent N messages, and collapse the rest as "N messages not analyzed"
im push failed	Output only to the conversation, and tell the user to send it manually
docs save failed	Output only Markdown to the conversation
Message content is an image/file	Mark "[image/file]" in the daily report, and suggest the user check it manually

Permission Table

Command	Required scope
im +chat-search	im:chat:readonly
im +chat-messages-list	im:message:readonly
im +messages-send	im:message
docs +create	docx:document:create

References

• ../lark-im/SKILL.md — IM atomic operations
• ../lark-doc/SKILL.md — document creation