Automator

Automator Skill

Build profitable automation workflows that save hours every week

When to Use

✅ USE this skill when:

When NOT to Use

❌ DON'T use this skill when:

💰 Value Proposition

What you get:

ROI Example:

Core Concepts

Workflow Structure

workflow:
  name: "Daily Report Generator"
  schedule: "0 8 * * *"  # Every day at 8 AM
  steps:
    - id: fetch_data
      task: "Fetch sales data from API"
      agent: "data-fetcher"
      timeout: 300
    - id: process
      task: "Process data into report format"
      agent: "data-processor"
      depends_on: [fetch_data]
      timeout: 600
    - id: notify
      task: "Send report via email"
      agent: "notifier"
      depends_on: [process]
      timeout: 120

Agent Roles

Each step can run on a specialized agent:

Error Handling & Retries

retry_policy:
  max_attempts: 3
  backoff: "exponential"  # 1s, 2s, 4s
  on_failure: "notify_admin"  # or "continue", "abort"
failure_notifications:
  - email: "admin@company.com"
  - slack: "#alerts"

Quick Start

1. Define Your Workflow

Create a YAML file my-workflow.yaml:

workflow:
  name: "Price Monitor"
  description: "Check product prices hourly and alert if below threshold"
  schedule:
    type: "interval"
    every: "1h"
steps:
  - name: "Check Amazon Price"
    agent: "price-checker"
    prompt: |
      Check price of product https://amazon.com/dp/B08XYZ
      Return price and availability
  - name: "Compare to Threshold"
    agent: "decision-maker"
    prompt: |
      Threshold: $50
      Current price: {{Check Amazon Price.output}}
      Is price below threshold? Return yes/no
  - name: "Send Alert if Cheap"
    agent: "notifier"
    prompt: |
      If {{Compare to Threshold.output}} == "yes":
        Send email to user@example.com
        Subject: Price Alert!
        Body: Product is now ${{Check Amazon Price.output}}
    depends_on: [Check Amazon Price, Compare to Threshold]

2. Load and Start

# Load workflow definition
openclaw workflow load my-workflow.yaml
Start the scheduled workflow
openclaw workflow start Price Monitor
Check status
openclaw workflow status

3. Monitor Execution

# View recent runs
openclaw workflow runs Price Monitor --limit 10
Get execution details
openclaw workflow run 
Stop workflow
openclaw workflow stop Price Monitor

Common Workflow Patterns

Pattern 1: Data Pipeline

workflow:
  name: "Daily Analytics Pipeline"
  schedule: "0 6 * * *"  # 6 AM daily
steps:
  - fetch: "Extract data from 3 sources"
    agent: "extractor"
    parallel: true  # Run multiple sources in parallel
  - transform: "Clean and normalize data"
    agent: "transformer"
    depends_on: [fetch]
  - analyze: "Generate insights"
    agent: "analyst"
    depends_on: [transform]
  - report: "Create PDF report"
    agent: "reporter"
    depends_on: [analyze]
  - distribute: "Email and Slack"
    agent: "distributor"
    depends_on: [report]

Benefit: 30-minute manual process → fully automated

Pattern 2: Approval Workflow

workflow:
  name: "Document Approval"
  trigger: "manual"  # Start on demand
steps:
  - draft: "Generate initial document"
    agent: "writer"
  - review: "Human review"
    agent: "approver"
    type: "human_input"  # Waits for manual approval
  - finalize: "Apply final changes"
    agent: "editor"
    depends_on: [review]
  - publish: "Deploy to production"
    agent: "publisher"
    depends_on: [finalize]

Benefit: Track approvals, no lost emails

Pattern 3: Alert & Escalation

workflow:
  name: "System Monitor"
  schedule: "*/5 * * * *"  # Every 5 minutes
  alert_levels:
    - warning: "System load > 80%"
    - critical: "System load > 95%"
steps:
  - check: "Monitor system metrics"
    agent: "monitor"
  - classify: "Determine severity"
    agent: "classifier"
  - alert:
      agent: "alerter"
      escalation:
        warning: "log_only"
        critical: ["slack", "pagerduty", "sms"]

Benefit: 24/7 monitoring without human attention

Advanced Features

Parallel Execution

steps:
  - name: "Parallel Fetch"
    agent: "fetcher"
    task: "Fetch data from multiple sources"
    parallel: true
    max_concurrent: 5

Conditional Branching

steps:
  - validate: "Check data quality"
    agent: "validator"
  - if_good:
      agent: "loader"
      depends_on: [validate]
      condition: "{{validate.output}} == 'valid'"
  - if_bad:
      agent: "alerter"
      depends_on: [validate]
      condition: "{{validate.output}} != 'valid'"

Output Passing

Use {{step-name.output}} to reference previous step results:

steps:
  - fetch_users:
      agent: "query-db"
      output: "user_ids"
  - fetch_data:
      agent: "api-client"
      prompt: "Fetch records for users: {{fetch_users.output}}"

Pro Tips

1. Start Simple, Then Complex

2. Use Specialized Agents

3. Implement Checkpoints

steps:
  - step1: ...
  - checkpoint: "Save progress to DB"
    agent: "checkpointer"
  - step2:
      depends_on: [checkpoint]
      # Will resume from checkpoint if failed

4. Set Up Alerts

5. Monitor Costs

Templates

Copy these templates to get started:

Template: Daily Summary

workflow:
  name: "Daily Digest"
  schedule: "0 7 * * *"
steps:
  - news: "Fetch latest news"
    agent: "news-fetcher"
  - weather: "Get weather forecast"
    agent: "weather-checker"
  - calendar: "Today's meetings"
    agent: "calendar-agent"
  - compile: "Compile into digest"
    agent: "compiler"
  - send: "Email digest"
    agent: "emailer"

Template: E-commerce Monitor

workflow:
  name: "Store Monitor"
  schedule: "*/15 * * * *"
steps:
  - check_inventory:
      agent: "inventory-checker"
      prompt: "List products below reorder threshold"
  - check_orders:
      agent: "order-checker"
      prompt: "Find pending orders > 24 hours"
  - generate_report:
      agent: "reporter"
      depends_on: [check_inventory, check_orders]
  - notify_manager:
      agent: "slack-notifier"
      depends_on: [generate_report]

Troubleshooting

Workflow Not Running?

Steps Timing Out?

No Output Available?

Want to Pause?

openclaw workflow pause 
openclaw workflow resume 

Next Steps

1. 📖 Read examples: ~/.openclaw/workspace/skills/automator/examples/ 2. 🧪 Test in sandbox: Use non-production agents first 3. 📈 Monitor usage: Check token costs daily 4. 🔄 Iterate: Refine prompts based on results 5. 📤 Share: Publish your workflows to ClawHub (coming soon!)

💡 Need Help?

_Automate the boring stuff. Focus on what matters._ 🚀

1. Define Your Workflow

Create a YAML file my-workflow.yaml:

workflow:
  name: "Price Monitor"
  description: "Check product prices hourly and alert if below threshold"
  schedule:
    type: "interval"
    every: "1h"
steps:
  - name: "Check Amazon Price"
    agent: "price-checker"
    prompt: |
      Check price of product https://amazon.com/dp/B08XYZ
      Return price and availability
  - name: "Compare to Threshold"
    agent: "decision-maker"
    prompt: |
      Threshold: $50
      Current price: {{Check Amazon Price.output}}
      Is price below threshold? Return yes/no
  - name: "Send Alert if Cheap"
    agent: "notifier"
    prompt: |
      If {{Compare to Threshold.output}} == "yes":
        Send email to user@example.com
        Subject: Price Alert!
        Body: Product is now ${{Check Amazon Price.output}}
    depends_on: [Check Amazon Price, Compare to Threshold]

2. Load and Start

# Load workflow definition
openclaw workflow load my-workflow.yaml
Start the scheduled workflow
openclaw workflow start Price Monitor
Check status
openclaw workflow status

3. Monitor Execution

# View recent runs
openclaw workflow runs Price Monitor --limit 10
Get execution details
openclaw workflow run 
Stop workflow
openclaw workflow stop Price Monitor

Workflow Not Running?

Steps Timing Out?

No Output Available?

Want to Pause?

openclaw workflow pause 
openclaw workflow resume