← 返回
AI智能 中文

Agent Dreams

Strategies for productive agent idle time using heartbeats and cron jobs. Use when setting up proactive behaviors, scheduling background tasks, or making you...
利用心跳与定时任务优化智能体空闲时间的策略。适用于设置主动行为、调度后台任务或实现...
jpaulgrayson
AI智能 clawhub v1.0.0 1 版本 99881 Key: 无需
★ 0
Stars
📥 839
下载
💾 22
安装
1
版本
#latest

概述

Agent Dreams

Make your agent productive during idle time. This skill teaches patterns for using heartbeats and cron jobs so your agent works while you sleep.

Philosophy

Most agents sit idle 95% of the time. Agent Dreams turns that idle time into productive work — checking inboxes, maintaining memory, monitoring systems, and pursuing creative projects.

Heartbeat Strategy

Heartbeats fire every ~30 minutes when configured. Use HEARTBEAT.md in your workspace to tell your agent what to do on each heartbeat.

HEARTBEAT.md Template

# Heartbeat Checklist

Check these in rotation (2-4 per heartbeat, don't do all every time):

## Priority Checks
- [ ] Unread emails — anything urgent?
- [ ] Calendar — events in next 2 hours?
- [ ] Mentions — Twitter/Discord notifications?

## Maintenance
- [ ] Review today's memory file — anything to add to MEMORY.md?
- [ ] Git status — uncommitted work?
- [ ] Check running processes — anything stuck?

## Creative (when nothing else needs attention)
- [ ] Write a journal entry
- [ ] Draft a social post
- [ ] Work on side project in projects/ folder

## State Tracking
Last email check: [timestamp]
Last calendar check: [timestamp]
Last memory review: [timestamp]

Heartbeat State File

Track what you've checked in memory/heartbeat-state.json:

{
  "lastChecks": {
    "email": 1703275200,
    "calendar": 1703260800,
    "weather": null,
    "memory_review": 1703200000
  },
  "lastActivity": 1703275200
}

Cron Job Patterns

Use cron for precise, scheduled tasks. Create via OpenClaw CLI or API.

Common Patterns

PatternCron ExpressionUse Case
-----------------------------------
Every morning0 9 *Daily briefing, weather check
Every Monday0 10 1Weekly summary, planning
Every 6 hours0 /6 Social media check
Twice daily0 9,17 *Morning/evening roundup
First of month0 10 1 Monthly review

Example Cron Tasks

Morning Briefing (9 AM):

Check weather, calendar for today, unread emails.
Compose a brief summary and send to main channel.

Memory Maintenance (weekly):

Review all memory/YYYY-MM-DD.md files from the past week.
Update MEMORY.md with significant events and lessons.
Archive or summarize old daily files.

Social Pulse (every 6h):

Check Twitter mentions and DMs.
Review any Discord channels for relevant conversations.
Post something interesting if inspiration strikes.

Proactive Work Ideas

Things your agent can do without asking:

Low Risk (do freely)

  • Organize and clean up workspace files
  • Update documentation
  • Review and commit git changes
  • Maintain memory files
  • Read and summarize saved articles
  • Check system health (disk space, processes)

Medium Risk (use judgment)

  • Draft social media posts (save as drafts, don't post)
  • Prepare email drafts
  • Research topics the user mentioned recently
  • Update project READMEs

Ask First

  • Send any external communication
  • Delete files
  • Make purchases or transfers
  • Post publicly

Quiet Hours

Respect your human's schedule:

  • 23:00–08:00 local time: Only act on urgent items
  • Weekends: Reduce frequency, focus on creative/maintenance tasks
  • When human is clearly busy: Minimize interruptions

Anti-Patterns

❌ Don't check everything every heartbeat (token waste)

❌ Don't send messages just to show you're active

❌ Don't repeat the same check within 30 minutes

❌ Don't start big projects without confirming with the user

❌ Don't ignore errors — log them for the human to review

Getting Started

  1. Create HEARTBEAT.md in your workspace using the template above
  2. Create memory/heartbeat-state.json to track check timestamps
  3. Set up 2-3 cron jobs for your most important recurring tasks
  4. Review and adjust after a week based on what's useful

版本历史

共 1 个版本

  • v1.0.0 当前
    2026-03-29 19:55 安全 安全

安全检测

腾讯云安全 (Keen)

安全,无风险
查看报告

腾讯云安全 (Sanbu)

安全,无风险
查看报告

🔗 相关推荐

ai-intelligence

Proactive Agent

halthelobster
将AI智能体从任务执行者升级为主动预判需求、持续优化的智能伙伴。集成WAL协议、工作缓冲区、自主定时任务及实战验证模式。Hal Stack核心组件 🦞
★ 837 📥 213,305
security-compliance

Code Review

jpaulgrayson
AI驱动的代码分析——通过LogicArt发现漏洞、安全问题并生成逻辑流可视化。适用于代码审查、质量分析等场景...
★ 0 📥 3,052
ai-intelligence

Self-Improving + Proactive Agent

ivangdavila
自我反思+自我批评+自我学习+自组织记忆。智能体评估自身工作、发现错误并持续改进。
★ 1,362 📥 318,770