SKILL.md
Ark DevSpace
Run Ark from a cloned repository using devspace for development/PR testing.
CRITICAL: Failure Behavior
If devspace fails, you MUST STOP and report the error. Do NOT try alternative approaches like ark-cli or npm install. DevSpace failures must be diagnosed and fixed.
When to use
- Testing Ark pull requests
- Running Ark from source code
Prerequisites
- Ark repo cloned (e.g.,
/workspace/arkor/workspace/agents-at-scale-ark) - Kind cluster running (use kind-setup skill if needed)
- Verify cluster:
kubectl cluster-info
Steps
- Navigate to ark repo
``bash cd /workspace/ark # or /workspace/agents-at-scale-ark ``
- Deploy with devspace (non-interactive)
``bash devspace deploy ``
This deploys all components and exits. Check the exit code - if non-zero, STOP and troubleshoot.
- Verify deployment succeeded
``bash kubectl get pods -n ark-system kubectl get pods -n cert-manager ``
- Wait for pods to be ready
``bash kubectl wait --for=condition=Ready pods --all -n ark-system --timeout=300s ``
- Check services
``bash kubectl get svc -n ark-system ``
Troubleshooting
Check devspace logs
ls -la .devspace/logs/
cat .devspace/logs/default.log
Common failures
- "namespace not found" - Set namespace first
``bash devspace use namespace ark-system ``
- "image pull error" - Build images locally
``bash devspace build ``
- Cert-manager not ready - Wait for it
``bash kubectl wait --for=condition=Available deployment --all -n cert-manager --timeout=120s ``
- Cluster connectivity - Re-export kubeconfig
``bash kind export kubeconfig --name ark-cluster --internal ``
If devspace exits with error
- Check the exit code and error message
- Check
.devspace/logs/for detailed logs - Check pod status:
kubectl get pods -A - Check events:
kubectl get events -A --sort-by='.lastTimestamp' - STOP and report the specific error - do not try workarounds
Cleanup
devspace purge
kubectl delete ns ark-system --ignore-not-found