[Creator] [design-spec] gcp-compute-vm-health

[rw-codebundle-agent]

Closes #152

Adds the gcp-compute-vm-health CodeBundle that monitors the health of standalone GCP Compute Engine VMs (instances NOT part of an instance group, which are covered by gcp-compute-instancegroup-health).

What it monitors

• Uptime / operational status — flags VMs running longer than UPTIME_WARNING_DAYS without reboot (severity 2) or in a non-RUNNING state (severity 3)
• OS patch status — OS Config vulnerability reports and OS policy compliance (severity 2–3)
• Disk utilization — disks above DISK_USAGE_THRESHOLD% (severity 4) or degraded (severity 3)
• Network health — IP assignment, tag ↔ firewall consistency, packet-loss indicators (severity 3)
• Guest / serial console health — boot, kernel, guest-agent, console error patterns (severity 2–3)
• Health summary — consolidated per-VM JSON summary with overall verdict (severity 3)

Implementation

• runbook.robot — 7 tasks (discover + 5 per-VM checks + project summary)
• sli.robot — in-repo 0–1 SLI averaging five health dimensions
• 7 bash scripts + gcp_vm_common.sh shared helpers (standalone-VM discovery excludes instance-group members)
• Generation rules + SLX / taskset / sli / slo templates (one SLX per standalone VM)
• .test/ Terraform fixtures (healthy VM, STOPPED unhealthy VM, grouped VM for exclusion testing)

Scoring

python -m scorer.score → 104/104 (PASS), threshold 70.

Dev notes

• Resource type uses the CloudQuery table name gcp_compute_instances (consistent with existing GCP bundles).
• Requires roles/compute.viewer, roles/osconfig.viewer, and roles/monitoring.viewer (for disk/network metrics) on the service account.