infrastructure-as-codelisted

# Purpose Enforce the invariant that a fresh server deploy of the project must succeed from code alone, without any manual `kubectl patch`, `kubectl exec`, or `helm --set` overrides that are not persisted. Covers config drift detection, the populated values pattern, post-incident IaC capture, and the "Where Changes Go" decision map. # When to use this skill (triggers) - Deploying the platform to a new server or environment and verifying all config is in code - Auditing config drift after manual cluster fixes (kubectl patch, psql exec, gcloud console changes) - Preparing or regenerating `.populated.yaml` values files via the infrastructure repo's `mapEnvValuesFromEnv.sh` (or equivalent env-mapping script) - Verifying that all manual cluster changes are captured in the correct source file (Helm values, Terraform, migration script) - Diagnosing literal `$VARIABLE_NAME` strings appearing in K8s Secrets # When NOT to use this skill (anti-triggers) - Writing or debugging Helm chart templates -- use `helm-chart-expert` - Checking migration safety or schema alignment -- use `migration-check` - Deploying the edge service (project.json → device) to edge devices or managing k3s -- use `kubernetes-deployment` - Managing Keycloak realm/client configuration -- use `keycloak` # Required environment (Runtime: .claude/skills/infrastructure-as-code/SKILL.md) - Tools/libraries: `helm` (v3.10+), `kubectl`, `terraform`, `diff`, `bash` - Scripts: the infrastructure repo's `files/mapEnvValue