k8s-agent-skills
v1.10.0
Published
Agent skills for Kubernetes cluster operations — Cilium, Talos, Flux, Rook-Ceph, CNPG, Gitea, Tekton, Cert-Manager, VictoriaMetrics, ZITADEL, Harbor, Higress, KServe, Kubeflow, MariaDB, Vector, ExternalDNS, External Secrets, Vault, Dragonfly, Flagger, Sea
Maintainers
Keywords
Readme
k8s-agent-skills
Agent skills for Kubernetes cluster operations tooling. Each skill is a self-contained SKILL.md designed for agentic AI tools (OpenCode, Claude Code, Codex, Cursor) that load skills for task-specific expertise.
Designed for: Cilium, Talos, Flux, Rook-Ceph, CNPG, ZITADEL, Gitea, Tekton, Cert-Manager, VictoriaMetrics, Harbor, Higress, KServe, Kubeflow, Vault, ExternalDNS, External Secrets, Dragonfly, Flagger, MariaDB, Vector, Sealed Secrets, Stakater Reloader, NVIDIA GPU, Atlas, Nextcloud, Tailscale and more.
Quick Start
As an npm dependency (recommended)
npm install --save-dev k8s-agent-skills
# Skills auto-link on install (postinstall script).
# Restart your AI agent — ready to use.
# Re-link after npm update:
npx skills-link --global
# Verify:
npx skills-link --global --checkFrom a git clone
git clone https://github.com/Aidas-dev/k8s-agent-skills.git
cd k8s-agent-skills
./bin/skills-link --global
# Or via npm script:
npm run skills-link -- --global
# Verify:
./bin/skills-link --global --checkSkills use symlinks, so content always reflects the latest source. npm update + automatic postinstall = skills stay current.
Per-agent linking
./bin/skills-link # ~/.agents/skills/ only (default)
./bin/skills-link --opencode # ~/.config/opencode/skills/
./bin/skills-link --claude # ~/.claude/skills/
./bin/skills-link --codex # ~/.codex/skills/
./bin/skills-link --cursor # ~/.cursor/skills/
./bin/skills-link --dry-run # preview without linkingSkills
| Skill | What it covers | CRDs | |-------|---------------|------| | atlas | Atlas Operator — DB schema migrations, lint, policies | AtlasSchema, AtlasMigration | | cert-manager | TLS cert provisioning, ACME, Issuers, Gateway integration | Certificate, Issuer, ClusterIssuer | | cilium-gateway | Gateway API, TLS, traffic splitting, oauth2-proxy, hostNetwork | GatewayClass, Gateway, HTTPRoute, etc. | | cilium-network | Cilium CNI, network policies, LB IPAM, encryption, Hubble | CiliumNetworkPolicy, CiliumCIDRGroup, etc. | | cnpg | CloudNativePG — PostgreSQL clusters, backups, poolers | Cluster, Backup, ScheduledBackup, Pooler | | dragonfly | DragonflyDB — Redis-compatible operator, replication, TLS | Dragonfly | | external-dns | DNS sync — Cloudflare, Route53, Gateway API, sources, registry | None | | external-secrets | External Secrets Operator — sync K8s Secrets from Vault, AWS, GCP, Azure | SecretStore, ClusterSecretStore, ExternalSecret, PushSecret | | flagger | Progressive delivery, canary, A/B, blue/green | Canary, MetricTemplate, AlertProvider | | flux | Flux CD router — debugging, CRDs, repo audit | Router → sub-skills | | helm | Helm router — chart development, production operations, OCI, GitOps, release inspection | Router → sub-skills | | helm-chart | Writing Helm charts — anatomy, templates, Go syntax, helpers, values, schema, testing | None | | helm-ops | Helm operations — release inspection, OCI, hooks, GitOps, troubleshooting | None | | gitea | Gitea router — API, runner, registry, webhooks, tea CLI | Router → sub-skills | | gitea-api | Gitea REST API — auth, repos, issues, PRs, packages | None | | gitea-registry | Gitea container registry — OCI, multi-arch, push/pull | None | | gitea-runner | Gitea Actions runners — registration, host-mode, ephemeral | None | | gitea-tea | tea CLI — commands, auth, actions, webhooks, admin | None | | gitea-webhooks | Gitea webhooks — events, HMAC, org vs repo hooks | None | | harbor | Harbor router — API, Helm, Terraform | Router → sub-skills | | harbor-api | Harbor REST API v2 — projects, artifacts, robots, replication, GC, OIDC | None | | harbor-helm | Harbor Helm chart — production deploy, external DB/Redis/S3, Trivy | None | | harbor-terraform | Harbor Terraform provider — 20 resources, 8 data sources | None | | higress | Higress router — CRDs, Wasm plugins, AI Gateway, Helm | Router → sub-skills | | higress-helm | Higress Helm chart — core/console/redis/plugin-server/o11y | None | | higress-operator | Higress CRDs — WasmPlugin, Http2Rpc, McpBridge, 41 Wasm plugins, 16 AI providers | WasmPlugin, Http2Rpc, McpBridge | | kserve | KServe router — CRDs, Helm, deployment modes | Router → sub-skills | | kserve-helm | KServe Helm — 10 charts, Serverless/Raw/ModelMesh modes | None | | kserve-operator | KServe CRDs — InferenceService, ServingRuntime, InferenceGraph, LLMInferenceService, LocalModel | 22 CRDs under serving.kserve.io | | kubeflow | Kubeflow router — Trainer v2, Pipelines v2, Training Operator v1 | Router → sub-skills | | kubeflow-pipelines | KFP v2 SDK — DSL, IR YAML, control flow, Kubernetes Native API | PipelineRun (v2beta1) | | kubeflow-trainer | Kubeflow Trainer v2.2 — TrainJob, TrainingRuntime, 5 ML policies | TrainJob, TrainingRuntime, ClusterTrainingRuntime | | kubeflow-training-operator | Legacy v1 — PyTorchJob, TFJob, MPIJob, XGBoostJob | PyTorchJob, TFJob, MPIJob, XGBoostJob | | mariadb | MariaDB router — operator CRDs, Helm | Router → sub-skills | | mariadb-helm | MariaDB operator Helm — 3 charts, production HA values | None | | mariadb-operator | MariaDB operator — 12 CRDs, Galera HA, MaxScale, backups, PITR | 12 CRDs under k8s.mariadb.com | | nextcloud | Nextcloud router — Helm deployment, Provisioning API | Router → sub-skills | | nextcloud-helm | Nextcloud Helm chart — NC34, S3, Imaginary, Collabora, cron | None | | nextcloud-api | Nextcloud Provisioning API — users, groups, apps | None | | netbird | NetBird router — server, client CLI, Terraform, K8s operator | Router → sub-skills | | netbird-cli | NetBird client — up/down, status, networks, expose, SSH, k8s | None | | netbird-server | NetBird control plane — helmforge chart, config.yaml, DB, Dex IdP, gRPC routing | None | | netbird-terraform | NetBird Terraform — groups, peers, routes, networks, policies, setup keys, DNS | None | | netbird-operator | NetBird K8s operator — SetupKey, NetworkRouter, NetworkResource, ClusterProxy | 7 CRDs under netbird.io/v1alpha1 | | nvidia-device-plugin | GPU discovery, GFD, NFD, CDI, MIG, time-slicing | None (ConfigMap) | | rook-ceph | Rook-Ceph router — operator CRDs, toolbox CLI | Router → sub-skills | | rook-ceph-operator | Ceph cluster, block pools, object store, NFS, CSI | CephCluster, CephBlockPool, CephObjectStore, etc. | | rook-ceph-toolbox | Ceph CLI — health, OSD mgmt, RBD, RGW, CRUSH | None (toolbox ops) | | sealed-secrets | Encrypted Secrets for GitOps, kubeseal, key rotation | SealedSecret | | stakater-reloader | ConfigMap/Secret reload, annotations, Helm values | None (annotation-based) | | tailscale | Tailscale router — CLI, status, ping, Talos integration | Router → sub-skills | | tailscale-talos | Tailscale on Talos — system extension, machine config, subnet routes | None | | talos | Talos Linux router — CLI ops, machine config, Terraform, tuppr upgrades | Router → sub-skills | | talos-terraform | Talos Terraform provider — machine secrets, config, cluster, upgrades | None | | talosctl | talosctl CLI — cluster lifecycle, upgrades, diagnostics, etcd | None | | talosconfig | Talos machine config — multi-document YAML, generation, patching | None | | tuppr | tuppr upgrade controller — TalosUpgrade/KubernetesUpgrade CRs, health-gated rollouts | TalosUpgrade, KubernetesUpgrade | | tekton | Tekton router — pipeline authoring, operator, Pipelines-as-Code | Router → sub-skills | | tekton-operator | Tekton operator — TektonConfig, profiles, pruner, Results DB | TektonConfig, TektonPipeline, etc. (operator.tekton.dev) | | tekton-pac | Pipelines-as-Code — .tekton/, Repository CR, ChatOps, tkn pac | Repository (pipelinesascode.tekton.dev) | | tekton-pipelines | Tekton pipeline authoring — resolvers, matrix, when/CEL, podTemplate | None | | vault | Vault router — Helm, API, Terraform | Router → sub-skills | | vault-api | Vault REST API — health, init, unseal, auth, KV, policies | None | | vault-helm | Vault Helm chart — HA+Raft, injector, storage, TLS, telemetry | None | | vault-terraform | Vault Terraform provider — 188+ resources, auth backends, KV, policies, identity | None | | vector | Vector router — Helm, operator CRDs | Router → sub-skills | | vector-helm | Vector Helm chart — 3 roles (Agent/Aggregator/Stateless), customConfig | None | | vector-operator | Vector operator — 5 CRDs, auto-routing by source type | Vector, VectorPipeline, ClusterVectorPipeline, VectorAggregator, ClusterVectorAggregator | | victoria-metrics | VM skill router — operator, queries, cardinality, logs, traces | Router | | victoriametrics-operator | VM Operator CRDs — VMAgent, VMAlert, VMServiceScrape, VLogs | 19 CRDs | | zitadel | ZITADEL router — API, Helm, Terraform | Router → sub-skills | | zitadel-api | ZITADEL API — OIDC/SAML/API apps, users, orgs, roles | None | | zitadel-helm | ZITADEL Helm — CNPG, Gateway API, caches, masterkey | None | | zitadel-terraform | ZITADEL Terraform provider — 80+ resources, 40+ data sources | None |
How It Works
Skills are symlinked from the npm package to your agent's skill directory:
node_modules/k8s-agent-skills/skills/cilium-network/SKILL.md
→ ~/.agents/skills/cilium-network/SKILL.md (OpenCode)
→ ~/.config/opencode/skills/cilium-network/ (OpenCode config)
→ ~/.claude/skills/cilium-network/ (Claude Code)
→ ~/.codex/skills/cilium-network/ (Codex CLI)
→ ~/.cursor/skills/cilium-network/ (Cursor)This means:
- Skill content updates instantly on
npm update— symlinks are live - New skills auto-link via
postinstallscript on package install - Stale symlinks (skills removed from a release) are cleaned automatically
- Run
./bin/skills-link --global --check(clone) ornpx skills-link --global --check(installed) to verify
Manual Installation
git clone https://github.com/Aidas-dev/k8s-agent-skills.git
cd k8s-agent-skills
./bin/skills-link --globalnpx skills-link only works when the package is installed as a dependency (node_modules/.bin/skills-link exists). From a clone, use ./bin/skills-link or npm run skills-link.
npm Publishing
Auto-publishes on v* tag push via GitHub Actions with OIDC Trusted Publisher — no tokens needed.
Easiest: GitHub UI — Actions → "Publish to npm" → Run workflow → pick patch/minor/major. The workflow bumps the version, creates the tag, and the tag push publishes + creates the GitHub Release.
Or locally:
npm version patch # or minor / major — creates v-prefixed tag automatically
git push origin main --tags
git push github main --tagsAlways use npm version, never git tag directly. It ensures the tag points to the right commit and uses the v prefix. Don't add [skip ci] in the version message — it suppresses the publish workflow. The workflow also fails fast if the tag and package.json versions disagree. See CONTRIBUTING.md for details.
License
MIT — free to use, modify, and distribute.
