Introduction
Scheduled reminders are one of the most practical features of a personal AI assistant. Imagine casually saying "Remind me about the meeting tomorrow at 9 AM" in a chat window and actually getting a notification when the time comes — that's exactly what we'll build in this tutorial.
This tutorial walks you through developing a full-featured reminder Skill with natural language time parsing, recurring reminders, and persistent storage.
Feature Planning
Our reminder Skill needs the following capabilities:
| Feature | Description | Priority |
|---|---|---|
| Set Reminders | Create one-time reminders using natural language | Required |
| Recurring Reminders | Daily, weekly, and monthly repeating reminders | Required |
| View Reminders | List all pending reminders | Required |
| Delete Reminders | Cancel a specific reminder | Required |
| Persistence | Reminders survive restarts | Required |
| Multi-Channel Push | Deliver notifications in the channel where the reminder was set | Recommended |
MCP Tool Configuration
The reminder feature requires file read/write and scheduling capabilities. Configure the MCP tools in ~/.config/openclaw/openclaw.json5:
{
mcp: {
servers: {
"filesystem": {
command: "npx",
args: ["-y", "@modelcontextprotocol/server-filesystem",
"--allow-read", "--allow-write",
"~/.openclaw/data/reminders/"],
},
"scheduler": {
command: "npx",
args: ["-y", "@openclaw-mcp/scheduler"],
}
}
}
}
Create the reminder data directory:
mkdir -p ~/.openclaw/data/reminders/
Writing the SKILL.md
Create reminder.SKILL.md in the ~/.openclaw/skills/ directory:
---
name: reminder
version: 1.0.0
description: 定时提醒技能，支持自然语言设置提醒、周期任务和持久化存储
triggers:
- 提醒
- remind
- 闹钟
- alarm
- 定时
- 待办
- 记得
- 别忘了
- 到时候
mcp_tools:
- filesystem
- scheduler
---
# 定时提醒技能
## Description
这是一个定时提醒技能，让用户可以通过自然语言设置、管理提醒事项。
提醒到期时，自动在用户所在的频道发送通知。
## Core Behavior
### 设置提醒
当用户想设置提醒时：
1. **解析时间**：从用户消息中提取时间信息，支持以下格式：
- 绝对时间："明天早上9点"、"3月15日下午3点"、"2026-03-20 14:00"
- 相对时间："5分钟后"、"1小时后"、"3天后"
- 自然语言："今天晚上"、"这周五"、"下周一上午"
2. **提取内容**：识别提醒内容，即用户想被提醒做什么。
3. **确认信息**：向用户确认提醒的时间和内容。
4. **保存提醒**：使用 filesystem 工具将提醒保存到
`~/.openclaw/data/reminders/reminders.json`。
5. **注册调度**：使用 scheduler 工具注册定时任务。
### 查看提醒
当用户要查看提醒列表时，读取 reminders.json 并按时间排序展示。
### 删除提醒
当用户要取消提醒时，根据编号或内容匹配找到提醒并删除。
### 触发提醒
当提醒时间到达时，scheduler 会回调 OpenClaw，在对应频道发送提醒消息。
## Data Schema
提醒数据使用 JSON 格式存储：
```json
{
"reminders": [
{
"id": "rem_20260328_001",
"content": "参加下午的产品会议",
"triggerTime": "2026-03-28T14:00:00+08:00",
"repeat": null,
"channelType": "telegram",
"channelId": "12345678",
"userId": "user_001",
"createdAt": "2026-03-28T09:00:00+08:00",
"status": "pending"
}
]
}
Field Reference
| Field | Type | Description |
|---|---|---|
| id | string | Unique identifier, format rem_YYYYMMDD_NNN |
| content | string | Reminder content |
| triggerTime | string | Trigger time in ISO 8601 format |
| repeat | object/null | Repeat rules (see below) |
| channelType | string | Source channel type |
| channelId | string | Source channel/user ID |
| userId | string | User who set the reminder |
| createdAt | string | Creation time |
| status | string | Status: pending/triggered/cancelled |
Repeat Rules
{
"repeat": {
"type": "daily",
"interval": 1,
"endDate": "2026-12-31T23:59:59+08:00"
}
}
Available type values:
daily: Repeat every dayweekly: Repeat every week (specific day of week can be specified)monthly: Repeat every month (specific date can be specified)cron: Custom cron expression
Output Format
Reminder Set Successfully
✅ 提醒已设置！
📋 内容：{提醒内容}
⏰ 时间：{YYYY年MM月DD日 HH:mm}
🔁 重复：{重复规则，无则显示"一次性"}
🆔 编号：{ID}
到时候我会在这里通知你。
View Reminder List
📝 你的提醒列表：
1. ⏰ {时间1} - {内容1} [编号: {ID1}]
2. ⏰ {时间2} - {内容2} [编号: {ID2}]
3. 🔁 {时间3} - {内容3}（每天重复）[编号: {ID3}]
共 {N} 个待触发提醒。
输入"取消提醒 {编号}"可删除指定提醒。
Reminder Triggered Notification
🔔 提醒时间到！
📋 {提醒内容}
⏰ 当前时间：{HH:mm}
---
这是你在 {设置时间} 设置的提醒。
Reminder Deleted Successfully
❌ 已取消提醒：{提醒内容}
Error Handling
- 如果无法解析时间，请用户用更明确的方式表述
- 如果时间已经过去，提醒用户设置的时间已过期
- 如果 reminders.json 不存在，自动创建空文件
- 如果 scheduler 不可用，告知用户提醒功能暂时不可用
Example Interactions
用户：明天早上8点提醒我买咖啡 助手：✅ 提醒已设置！ 📋 内容：买咖啡 ⏰ 时间：2026年03月29日 08:00 🔁 重复：一次性 🆔 编号：rem_20260328_001
用户：每周五下午5点提醒我写周报 助手：✅ 提醒已设置！ 📋 内容：写周报 ⏰ 时间：每周五 17:00（下次：2026年03月28日） 🔁 重复：每周五 🆔 编号：rem_20260328_002
用户：看看我有哪些提醒 助手：📝 你的提醒列表： 1. ⏰ 03/29 08:00 - 买咖啡 [编号: rem_20260328_001] 2. 🔁 每周五 17:00 - 写周报 [编号: rem_20260328_002] 共 2 个待触发提醒。
用户：取消买咖啡的提醒 助手：❌ 已取消提醒：买咖啡
## Persistence Mechanism
### Storage Strategy
Reminder data is saved in `~/.openclaw/data/reminders/reminders.json`. The file is written immediately after every operation (create, delete, or modify).
### Recovery After Restart
When OpenClaw restarts, the reminder skill needs to:
1. Read all reminders with a `pending` status from reminders.json
2. Re-register them with the scheduler
3. Clean up expired but untriggered reminders (mark them as `missed` and notify the user during the next interaction)
## Timezone Handling
Time parsing and storage must account for timezones:
```markdown
## Timezone Rules
- 默认使用用户所在频道配置的时区
- 如果未配置时区，默认使用 Asia/Shanghai (UTC+8)
- 所有存储时间使用 ISO 8601 带时区格式
- 显示时间使用用户本地时区
Testing Checklist
After development, test using the following checklist:
# 1. Restart to load the skill
openclaw restart
# 2. Confirm the skill is loaded
openclaw skill list
Then test each scenario in chat:
- [ ] Set a reminder for 1 minute from now and verify it triggers on time
- [ ] Set a reminder with an absolute time
- [ ] Set a recurring reminder
- [ ] View the reminder list
- [ ] Delete a reminder
- [ ] Restart OpenClaw and check that reminders are recovered
Summary
The scheduled reminder Skill is a comprehensive project that involves natural language time parsing, data persistence, scheduled task dispatching, and multi-channel notification delivery. Once you've mastered the development approach for this Skill, you'll be equipped to build even more complex automation workflow skills.