aether

Interactively investigate Aether outages, alerts, regressions, performance issues, failed deployments, missing telemetry, backup failures, authentication problems, and network/security questions using live read-only evidence, including human follow-up on an Inquest incident. Use Grafana as the primary correlation surface across Prometheus metrics, Loki logs, Tempo traces, ClickHouse Zeek/Suricata data, Kubernetes state, Fleet, Talos, SSH, and service APIs. This is not the unattended Grafana-to-Kestra-to-Holmes-to-GitLab pipeline owned by sibling ../inquest; consume its incident record when present, independently verify its RCA, and stop before remediation or live mutation.

Map an Aether component, service, host, or proposed change to its runtime, placement, authoritative IaC, configuration owner, Taskfile workflow, dependencies, observability, and current documentation. Use for architecture questions, locating source files, determining whether something runs in Kubernetes or on a VM/LXC/host/cloud, planning where a change belongs, onboarding to the repo, and resolving conflicts between docs and code. For Inquest, separate the sibling repo's automated alert flows and incident lifecycle from Aether's Kestra, Holmes, Grafana, secret, and network platform ownership. This skill orients; use investigate-aether for live incident evidence.

GitLab REST API and git workflows on gitlab.home.shdr.ch — search projects, clone, branch, push, open/update MRs, check pipelines.