Phase 7 / Ep 34: Agent Custom Workflow —— Enabling Agents to Define Their Own Working Style

⏱ Est. reading time: 3 min Updated on 4/13/2026

🎯 Learning Objective: Enable Agents to autonomously create and manage workflows.

1. Cron Scheduled Tasks

Enable Agents to execute tasks according to a schedule:

# Send daily report at 8 AM every day
openclaw cron add "Daily Briefing" --schedule "0 8 * * *" --action "Summarize yesterday's important events and send via Telegram"

# Check server status every hour
openclaw cron add "Server Inspection" --schedule "0 * * * *" --agent ops-agent --action "Check CPU, memory, disk usage"

# Generate weekly report every Monday
openclaw cron add "Weekly Report" --schedule "0 9 * * 1" --action "Organize all conversation summaries from this week and generate a weekly report"

# List all scheduled tasks
openclaw cron list

# Remove a task
openclaw cron remove "Daily Briefing"

2. Agent Collaboration

Define collaboration rules via BOOTSTRAP.md:

# Collaboration Rules
When the user requests research:
1. Forward the task to research-agent
2. Wait for research-agent to return results
3. Organize and inform the user

When the user requests deployment:
1. First, have code-agent run tests
2. After tests pass, have ops-agent execute deployment
3. Notify the user once deployment is complete

sequenceDiagram
    actor User as 🧑 User
    participant PA as 🧑‍💼 Personal
    participant CA as 💻 Code Agent
    participant OA as 🔧 Ops Agent

    User->>PA: "Help me deploy the latest version"
    PA->>CA: "Please run tests to confirm the code is correct"
    CA->>CA: npm test → ✅ All passed
    CA->>PA: "Tests passed, ready for deployment"
    PA->>OA: "Please deploy the main branch to production"
    OA->>OA: docker compose up -d
    OA->>PA: "Deployment complete, service restarted"
    PA->>User: "✅ Deployment complete! Tests passed + service restarted"

3. Agent Self-Evolution

Agents can improve themselves based on feedback:

# BOOTSTRAP.md
## Self-Improvement
Every Sunday at 10 PM:
1. Review all conversations from this week
2. Identify recurring issues
3. Update common Q&A in MEMORY.md
4. If a certain Skill is frequently used, suggest installing a more efficient alternative version

Next Episode Preview: Ep 35, In-depth Session Analysis—The complete lifecycle of conversation creation, maintenance, compression, and archiving.