Agent Workflow Composer
Chain OpenClaw skills into automated, multi-step workflows. Define pipelines, set conditions, handle errors, and route outputs between skills.
Quick Start
# Create a new workflow
python3 scripts/workflow-composer.py create --name "daily-security" --steps "security-audit-pro:scan,memory-health-monitor:status"
# List all workflows
python3 scripts/workflow-composer.py list
# Run a workflow
python3 scripts/workflow-composer.py run --name "daily-security"
# Show workflow details
python3 scripts/workflow-composer.py show --name "daily-security"
# Validate workflow (check all skills exist, params valid)
python3 scripts/workflow-composer.py validate --name "daily-security"
# Show available skills for building workflows
python3 scripts/workflow-composer.py skills
# Delete a workflow
python3 scripts/workflow-composer.py delete --name "daily-security"
Commands
| Command | Purpose |
|---------|---------|
| create | Create a new workflow from skill chain |
| list | List all saved workflows |
| run | Execute a workflow |
| show | Show workflow steps and configuration |
| validate | Check workflow for errors (missing skills, bad params) |
| skills | List available skills that can be used in workflows |
| delete | Remove a saved workflow |
| templates | Show pre-built workflow templates |
Workflow Definition
Workflows chain skills with skill:command steps separated by commas:
step1_skill:command → step2_skill:command → step3_skill:command
Each step can:
- Pass its output to the next step via temp files
- Stop the chain on error (fail-fast)
- Continue on error (resilient mode)
- Log results to workflow history
Pre-Built Templates
- daily-audit: security-audit-pro → memory-health-monitor → agent-cost-dashboard
- full-scan: skill-security-scanner → plugin-approval-guard → skill-dependency-mapper
- memory-maintenance: memory-compressor:analyze → memory-health-monitor:status
- cost-review: agent-cost-dashboard:summary → multi-model-evaluator:rank