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 record today's lunch" / "Record diet"
- "My stomach feels uncomfortable today" / "I didn't sleep well last night"
- "What did I eat this week?" / "Diet review"
- "Help me analyze the correlation between diet and physical condition"
- "Generate a weekly health report"
Prerequisites
Only supports user identity. Before execution, ensure authorization has been granted:
lark-cli auth login --domain im,base,docs
Workflow Overview
记录饮食：
用户发照片/描述 ──► AI 识别食物 ──► base +record-upsert
记录身体状态：
用户描述状态 ──► base +record-upsert
关联分析：
base +data-query ──► AI 分析饮食-状态关联
周报生成：
base +record-list ──► AI 汇总 ──► doc +create
主动提醒：
base +record-list ──► 识别风险食物 ──► im +messages-send
Data Initialization
Base Table Structure
Table 1: Diet Log (diet_log)
| Field Name | Type | Description |
|---|---|---|
| Date | Date | Record date |
| Meal | Single select | Breakfast/Lunch/Dinner/Snack/Late-night snack |
| Time | Date | Specific time |
| Food List | Text | Recognized food names (comma-separated) |
| Food Details | Text | Detailed description and portion size of each food |
| Photo Link | URL | Link to the food photo in Feishu |
| Estimated Calories | Number | Estimated calories (optional, AI estimate) |
| Notes | Text | Additional notes from the user |
| Created Time | Date | Record creation time |
Table 2: Body Status (body_status)
| Field Name | Type | Description |
|---|---|---|
| Date | Date | Status date |
| Time | Date | Specific time |
| Status Type | Single select | Stomach discomfort/Headache/Insomnia/Fatigue/Allergy/Skin/Mood/Other |
| Severity | Single select | Mild/Moderate/Severe |
| Status Description | Text | Detailed description |
| Possible Triggers | Text | Causes guessed by the user (if any) |
| Duration | Text | Duration of symptoms |
| Relief Methods | Text | How it was relieved (if any) |
| Created Time | Date | Record creation time |
Table 3: Correlation Analysis (correlation)
| Field Name | Type | Description |
|---|---|---|
| Food | Text | Suspicious food name |
| Status | Text | Associated body status |
| Occurrence count | Number | Number of times this combination occurred |
| Time difference pattern | Text | How long after eating symptoms appear (e.g. "2-4 hours") |
| Association strength | Single select | Low/Medium/High |
| Most recent occurrence date | Date | Most recent occurrence time |
| Notes | Text | Analysis notes |
Initialization command
# 注意：+table-create 可能部分成功，需捕获错误后用 +field-create 补字段
# 注意：连续创建字段会触发限流（错误码 800004135），每次调用间隔至少 1 秒
lark-cli base +table-create --name "饮食记录" --fields '[
{"name":"日期","type":"datetime"},
{"name":"餐次","type":"select","multiple":false,"options":[{"name":"早餐"},{"name":"午餐"},{"name":"晚餐"},{"name":"加餐"},{"name":"夜宵"}]},
{"name":"时间","type":"datetime"},
{"name":"食物列表","type":"text"},
{"name":"食物详情","type":"text"},
{"name":"照片链接","type":"link"},
{"name":"估算热量","type":"number"},
{"name":"备注","type":"text"},
{"name":"创建时间","type":"datetime"}
]'
lark-cli base +table-create --name "身体状态" --fields '[
{"name":"日期","type":"datetime"},
{"name":"时间","type":"datetime"},
{"name":"状态类型","type":"select","multiple":false,"options":[{"name":"胃部不适"},{"name":"头痛"},{"name":"失眠"},{"name":"疲劳"},{"name":"过敏"},{"name":"皮肤"},{"name":"情绪"},{"name":"其他"}]},
{"name":"严重程度","type":"select","multiple":false,"options":[{"name":"轻微"},{"name":"中等"},{"name":"严重"}]},
{"name":"状态描述","type":"text"},
{"name":"可能诱因","type":"text"},
{"name":"持续时间","type":"text"},
{"name":"缓解方式","type":"text"},
{"name":"创建时间","type":"datetime"}
]'
lark-cli base +table-create --name "关联分析" --fields '[
{"name":"食物","type":"text"},
{"name":"状态","type":"text"},
{"name":"出现次数","type":"number"},
{"name":"时间差模式","type":"text"},
{"name":"关联强度","type":"select","multiple":false,"options":[{"name":"低"},{"name":"中"},{"name":"高"}]},
{"name":"最近出现日期","type":"datetime"},
{"name":"备注","type":"text"}
]'
Mode 1: Record diet
Method A: Photo recognition
The user sends a food photo via Feishu message:
# 获取图片消息
lark-cli im +messages-resources-download --message-id "<message_id>" --file-key "<file_key>"
AI processing:
- Use multimodal capabilities to recognize the food in the photo
- Prompt the user to confirm or supplement: "Received, dinner is spaghetti bolognese pizza, how many slices did you eat?"
- After the user replies, record it in the diet record table
Method B: Text description
The user directly describes what they ate:
- "For lunch I ate rice, braised pork, and green vegetables"
- "In the afternoon I drank a latte"
AI processing:
- Parse food names
- If the description is vague, ask for details (e.g. "What was added to the coffee? Whole milk or oat milk?")
- Estimate calories (optional)
- Record in the diet record table
Record command
lark-cli base +record-upsert --table-id "<diet_table_id>" --json '{
"日期": "2026-03-31",
"餐次": "午餐",
"时间": "2026-03-31 12:30:00",
"食物列表": "米饭,红烧肉,青菜",
"食物详情": "米饭一碗、红烧肉约200g、炒青菜一份",
"备注": ""
}'
Mode 2: Record body status
The user describes their body status:
- "My stomach feels uncomfortable today" → Record: status type = stomach discomfort
- "I had insomnia last night, didn't fall asleep until around 3 a.m." → Record: status type = insomnia
- "I felt a bit dizzy after lunch" → Record: status type = headache + possible trigger = lunch
lark-cli base +record-upsert --table-id "<body_table_id>" --json '{
"日期": "2026-03-31",
"时间": "2026-03-31 14:00:00",
"状态类型": "胃部不适",
"严重程度": "中等",
"状态描述": "午饭后胃有点胀，持续约1小时",
"可能诱因": "午餐吃的太快"
}'
Mode 3: Association analysis
AI automatically analyzes the association between diet and body status:
Analysis logic
- Time window matching: Find diet records within 2-24 hours before the body status occurs
- Frequency statistics: Count the co-occurrence frequency of each food and each body status
- Time difference pattern: Analyze the pattern of the time interval from food consumption to symptom onset
- Exclude common factors: Exclude commonly occurring foods (such as rice, water), and focus on specific foods
Analysis command
# 获取最近 30 天的数据
lark-cli base +record-list --table-id "<diet_table_id>" --filter '...日期 >= "30天前"'
lark-cli base +record-list --table-id "<body_table_id>" --filter '...日期 >= "30天前"'
AI analysis output
## 饮食-身体状态关联分析（过去 30 天）
### 发现的关联模式
| 食物 | 身体状态 | 出现次数 | 典型时间差 | 关联强度 |
|------|---------|---------|-----------|---------|
| 冰美式 | 胃部不适 | 3次 | 1-2小时 | 高 |
| 辣味火锅 | 失眠 | 2次 | 6-8小时 | 中 |
### 建议
- 冰美式可能与你的胃部不适有较强关联，建议减少摄入或改用热饮
- 辣味火锅后出现失眠，建议避免晚餐食用辛辣食物
Update association analysis table
lark-cli base +record-upsert --table-id "<correlation_table_id>" --json '{
"食物": "冰美式",
"状态": "胃部不适",
"出现次数": 3,
"时间差模式": "1-2小时",
"关联强度": "高"
}'
Mode 4: Proactive reminders
When the user records a body status, AI proactively checks the association analysis table:
lark-cli base +record-list --table-id "<correlation_table_id>" --filter '...关联强度 IN ("高","中")'
If the current diet contains high-risk foods, remind immediately:
提醒：你今天喝了冰美式，之前 3 次喝冰美式后 1-2 小时内都出现了胃部不适。
建议：注意观察，如果出现不适请记录。
Mode 5: Weekly report generation
Generate a health report every week and save it as a Feishu document:
Report content
## 健康周报（{日期范围}）
### 饮食概览
- 记录天数：N 天
- 总餐次：N 次
- 饮食规律性：{评价}
### 身体状态概览
- 记录不适：N 次
- 主要症状：{列表}
- 严重程度分布：轻微 N 次、中等 N 次、严重 N 次
### 本周新发现
- {新发现的关联模式}
### 趋势对比（vs 上周）
- 饮食多样性：{变化}
- 不适频率：{变化}
### 建议
- {个性化建议}
Generation command
lark-cli doc +create --title "健康周报 ({日期范围})" --markdown "<报告内容>"
Permission table
| Command | Required scope |
|---|---|
im +chat-messages-list |
im:message:readonly |
im +messages-resources-download |
im:message.resource:readonly |
im +messages-send |
im:message:send_as_bot |
base +table-create |
bitable:app |
base +record-list |
bitable:app:read |
base +record-upsert |
bitable:app:write |
base +data-query |
bitable:app:read |
docs +create |
docx:document:create |
Reference
lark-shared— Authentication, permissions (required reading)lark-im—+chat-messages-list、+messages-resources-download、+messages-sendDetailed usagelark-base—+table-create、+record-list、+record-upsert、+data-queryDetailed usagelark-doc—+createDetailed usage