Agent Zero — OpenClaw Workspace Template

A fully configured AI agent workspace for OpenClaw, featuring persistent memory, structured organization, and automated workflows.

What is this?

This is the workspace template for Agent Zero, a personal AI operative built on OpenClaw. It demonstrates how to give an AI agent:

• Persistent memory across sessions (daily logs + curated long-term memory)
• PARA organization (Projects, Areas, Resources, Archives)
• Automated workflows via cron jobs (backups, monitoring, scanning)
• Custom tools for specific tasks
• A soul — personality, boundaries, and operating principles

Quick Start

1. Install OpenClaw
2. Clone this repo into your workspace: ~/.openclaw/workspace/
3. Edit USER.md with your info
4. Edit IDENTITY.md to name your agent
5. Customize SOUL.md to define personality
6. Start chatting — the agent reads these files every session

File Structure

├── AGENTS.md          # Operating manual — how the agent behaves
├── SOUL.md            # Personality, tone, boundaries
├── IDENTITY.md        # Agent name and identity (customize this)
├── USER.md            # About the human (customize this)
├── CONSTITUTION.md    # Rules and constraints
├── TOOLS.md           # Local tool notes (cameras, SSH, voices)
├── HEARTBEAT.md       # Proactive background task checklist
├── PARA.md            # Workspace organization map
├── MEMORY.md          # Long-term curated memory (starts empty)
├── memory/            # Daily logs (YYYY-MM-DD.md)
├── tools/             # Custom scripts and automation
│   ├── backup.sh      # GitHub backup script
│   └── memory-sync-check.sh  # Memory change detection
├── areas/             # Ongoing responsibilities
├── projects/          # Goal-oriented work with deadlines
├── resources/         # Reference material
└── archives/          # Completed/inactive items

Key Concepts

Memory System

The agent wakes up fresh every session. Continuity comes from files:

• memory/YYYY-MM-DD.md — Raw daily logs. What happened, what was discussed, decisions made.
• MEMORY.md — Curated long-term memory. The agent periodically reviews daily logs and distills important lessons, preferences, and context here.
• memory-sync-check.sh — Detects when MEMORY.md changes so the agent can re-index.

Each daily log ends with a ## Next Actions section — a prescriptive handoff to future sessions.

PARA Organization

Everything in the workspace follows the PARA method:

Heartbeats

OpenClaw sends periodic heartbeat polls. The agent uses these to:

• Check for new emails, calendar events, notifications
• Run background maintenance (memory review, git status)
• Proactively reach out when something needs attention

Configure what to check in HEARTBEAT.md.

Automated Backups

The included tools/backup.sh script:

• Syncs the full workspace to a private GitHub repo
• Includes sessions, cron jobs, memory, and agent configs
• Auto-tags weekly and monthly snapshots
• Prunes old commits monthly to keep repo size manageable

Set it up as a cron job:

openclaw cron add --name "daily-backup" \
  --schedule "0 23 * * *" --tz "America/Chicago" \
  --isolated --model sonnet \
  --message "Run: bash /path/to/tools/backup.sh 'Daily auto-backup'"

Cron Jobs

Example automated workflows you can build:

• Daily backups — Push workspace to GitHub nightly
• Security audits — Check permissions, exposed keys
• Newsletter monitoring — Fetch and summarize subscriptions
• Reminders — Time-based notifications via Telegram/Discord

See AGENTS.md for the cron job checklist (model tier, session type, delivery).

Customization

Make it yours

1. SOUL.md — This is the agent's personality. Make it match what you want.
2. IDENTITY.md — Name, emoji, vibe.
3. USER.md — Your name, timezone, preferences.
4. AGENTS.md — Operating rules. Add your own conventions.
5. CONSTITUTION.md — Hard constraints and safety rules.
6. TOOLS.md — Notes about your specific setup (devices, SSH hosts, etc).

Add tools

Drop scripts in tools/. The agent can run them via exec. Document usage in the script header or a companion .md file.

Memory tips

• Let the agent write daily logs — don't micromanage the format
• Periodically ask it to review and update MEMORY.md
• Use ## Next Actions in daily logs for session continuity
• MEMORY.md should be curated wisdom, not raw logs

Philosophy

You're not building a chatbot. You're building a right-hand operator.

The goal is an agent that:

• Remembers what matters across sessions
• Acts without being asked (via heartbeats and cron)
• Learns from mistakes (documented in MEMORY.md)
• Respects boundaries (asks before external actions)
• Has personality without being annoying

License

MIT — use it however you want.

Links

• OpenClaw
• OpenClaw Docs
• Community Discord
• ClawtHub Skills