Diagnose a broken/slow/crashlooping app in kubenuc or k8s-vms-daniele — gather evidence via Graylog/Grafana/kubectl before asking the user anything. Triggers on "X is down", "X is crashlooping", "X is erroring", "X is slow", "why did X restart".
Route documentation/notes to the right destination (Confluence, README/CLAUDE.md, Obsidian, or NetBox) based on content type, and confirm the exact destination with the user before writing anything. Also invoke proactively — before opening a PR — when the change is a new operational gotcha, an incident, a security-hardening trade-off, a version-pin rationale, a new app/cluster, or a structural Terraform/infra change, to decide whether README.md, a CLAUDE.md, or Confluence needs updating.
Cluster lifecycle operations — setting up new clusters, kubenuc-test overlay pattern, kustomize patch conventions, and FluxCD bootstrap structure for infra-cd clusters.
Before adding any new Prometheus scrape target or exporter, check the active-series budget in Grafana Cloud and propose write_relabel drops if needed. Triggers on "add a scrape config", "add an exporter", "enable metrics for X", "add a ServiceMonitor". Gates any change that could grow active series count.
Review a Renovate PR before merging — read the changelog, diff values against this repo's overrides, check e2e status, flag major bumps, and produce a merge checklist. Triggers on "review this renovate PR", "should I merge this chart bump", "is this renovate PR safe".
Ansible playbook/role conventions, Molecule testing patterns, and ansible-agent invocation for the infra-cd repository.
GitHub Actions workflow patterns for infra-cd — Terraform CI (fmt/init/validate/plan/apply), cluster validation (k3s + FluxCD + Robot Framework), Flux cron updates, and runner selection.
FluxCD/GitOps operations for infra-cd — add/update apps, HelmRepositories, SOPS variable substitution, and dependency chains. Delegates YAML generation to kubernetes-agent with Ollama.