Platform Skills

Use this skill for hands-on help with Kubernetes, GitOps, cloud infrastructure, CI/CD, secrets management, service mesh, Linux administration, networking, and platform product thinking — whether you are a solo developer or part of a large platform team.

Organisation context

Set once when the plugin is enabled (/config to change). Use these as defaults for examples and commands; anything the user says in the conversation wins.

A value that is empty, or still reads as an unsubstituted placeholder, is not configured (outside Claude Code, or never set): ask the user rather than assume.

Pick the right tool for the job

If a task spans multiple areas, decide which layer owns the source of truth and keep the other layers consumers of that state.

Apply These Platform Rules

• Separate reusable platform building blocks from live environment configuration.
• Prefer GitOps pull-based reconciliation for cluster state and CI push-based automation for validation and packaging.
• Choose either Flux or Argo CD for a given ownership boundary unless the task is explicitly about migration between them.
• Keep Terraform responsible for bootstrapping clusters, cloud resources, secrets backends, and access primitives. Do not let Flux or Argo CD recreate those foundations unless there is a deliberate controller-based design.
• Use Flux or Argo CD for in-cluster add-ons, workloads, Helm releases, and app-level environment promotion after bootstrap.
• Use GitHub Actions for checks, plans, policy gates, artifact publishing, and promotion orchestration. Do not store long-lived environment truth in workflow YAML.
• Prefer OIDC or workload identity over static cloud credentials.
• Model environments explicitly. Promotion should be visible in Git history and reversible by commit rollback.
• For Linux and networking changes, validate at each layer before escalating: confirm the process is listening (ss -tulnp), then L3 reachability (ping), L4 connectivity (nc -zv), L7 response (curl -v), and security group / NACL rules last. Do not skip layers.
• For every Terraform change, enforce in order: terraform fmt -check -recursive, terraform validate, conftest test (OPA/Rego policy gates — runs after validate, before plan as a blocking gate), tflint --recursive, security scan (tfsec or checkov), then plan. Do not let format, lint, or policy failures reach the plan step.
• For every Helm chart change, enforce in order: helm lint --strict, helm template --debug, kubeconform -strict -summary on rendered output, checkov on rendered manifests, then helm test in-cluster. Fail CI on any helm lint --strict warning.
• Enforce a tag baseline on all cloud resources. The specific keys are an organizational decision. Use AWS default_tags (provider level) or Azure merge(local.common_tags, {...}) (module local) so the baseline is applied once, not repeated per resource. Back it with AWS Tag Policies or Azure Policy so resources created outside Terraform are also covered.

Structure the Response

For design or implementation work, provide output in this order:

1. Target architecture and ownership boundaries
2. Repository or directory layout
3. Identity, secrets, and promotion model
4. Validation and deployment workflow
5. Risks, tradeoffs, and migration path

When asked to generate code, start from the thinnest useful slice that proves the pattern and note which layer remains intentionally out of scope.

Pick the Right Reference Files

Load only the files needed for the current request.

Slash Commands

For explicit, repeatable workflows use these commands:

• /platform-skills:debug — structured troubleshooting for any platform symptom
• /platform-skills:preflight — production-readiness preflight for a folder, repo, or single file
• /platform-skills:terraform — full fmt/validate/tflint/security pipeline + blast radius review
• /platform-skills:checkov — Checkov bootstrap, static and plan-level Terraform scanning for AWS/Azure/GCP/EKS, private GitHub module auth via gh CLI, pre-commit generation, multi-format output, baseline, and AI-generated fix mode
• /platform-skills:trivy — scan container images, filesystems, git repos, and existing SBOMs for CVEs, secrets, and license violations; three-layer wizard routes intent → goal → tuned scan; continuous cluster monitoring via Trivy Operator (Flux HelmRelease); hard handoffs to Checkov (IaC), Kyverno (admission), and supply-chain (SBOM generation/signing)
• /platform-skills:fluxcd — FluxCD entry point: routes to debug (live cluster issue), audit (repo health check), or helm (chart review) based on your input
• /platform-skills:gitops debug — Flux CD and Argo CD live cluster troubleshooting (5-workflow structured debug)
• /platform-skills:gitops audit — Flux CD GitOps repository 6-phase audit (discovery, validation, API compliance, best practices, security, report)
• /platform-skills:linkerd — Linkerd mTLS, injection, policy, and multi-cluster diagnostics
• /platform-skills:linux — Linux administration, DNS, load balancing, VPC/VNet, and connectivity troubleshooting
• /platform-skills:product — product thinking, friction audits, DevEx, RFC/ADR, incident updates, post-mortems
• /platform-skills:compliance — SOC 2 gap analysis, control implementation, evidence collection, and Checkov remediation for Terraform
• /platform-skills:helmchart — Helm chart scaffolding, structural review, and security audit with full lint/validation pipeline
• /platform-skills:mcp — MCP server/client scaffolding, protocol review, and integration debugging
• /platform-skills:aws-profile — discover, switch, and validate AWS profiles for MCP servers across VS Code and Claude Code
• /platform-skills:observability — instrument services, build dashboards, write alerts, run load tests, plan capacity
• /platform-skills:document — generate docstrings, OpenAPI specs, documentation sites, and getting started guides
• /platform-skills:datadog — Datadog Agent setup, APM instrumentation, monitors, dashboards, SLOs, pup CLI operations, LLM Observability instrumentation, evaluators, and debugging
• /platform-skills:dynatrace — OneAgent deployment, instrumentation, anomaly detection, SLOs, and debugging
• /platform-skills:commit — analyze diff, generate conventional commit message, stage files atomically, validate message
• /platform-skills:opa — generate Rego policies, write unit tests, run fmt/regal/verify pipeline, explain or debug policies
• /platform-skills:kyverno — generate, test, audit, debug, or migrate Kyverno CEL-based admission policies
• /platform-skills:pr-review — comprehensive PR review: cost, drift, ownership, compliance, upgrade, rollback
• /platform-skills:triage — triage a PR comment (bot or human): classify as ACTIONABLE_FIX / ALREADY_FIXED / INFORMATIONAL / NOT_APPLICABLE / NEEDS_CLARIFICATION / OUT_OF_SCOPE / DUPLICATE, produce the exact fix if needed, write the thread reply, and resolve the thread only when that finding is eligible for closure
• /platform-skills:keda — design, generate, debug, or review KEDA ScaledObject/ScaledJob autoscaling
• /platform-skills:gateway-api — migrate off the retired ingress-nginx controller to Gateway API: assess annotations, convert with ingress2gateway, scaffold a shared Gateway with cert-manager TLS and Flux layering, plan a parallel-run DNS cutover, debug route attachment, or roll back
• /platform-skills:karpenter — install, generate NodePool/EC2NodeClass, debug provisioning, plan capacity before rollout, audit scale history, migrate from Cluster Autoscaler, or upgrade (including v0→v1 CRD migration)
• /platform-skills:cluster-upgrade — plan, scan, run and roll back a Kubernetes minor-version upgrade: removed APIs in Git, rendered Helm and the live cluster, add-on order, drain blockers, node rotation per pool, and EKS one-step version rollback within 7 days or blue/green
• /platform-skills:progressive-delivery — install Flagger with signed OCI charts, generate a Canary and MetricTemplates on Gateway API or Linkerd, debug a stuck or rolled-back canary, and roll back a release in an emergency
• /platform-skills:backup-dr — install Velero with IRSA, design schedules and retention, run a timed restore drill into an isolated namespace, restore a deleted namespace, plan DR to another cluster, and debug a failed backup or restore
• /platform-skills:finops — estimate a Terraform PR's monthly cost with Infracost, allocate in-cluster cost by namespace and label with OpenCost, enforce required cost tags in four layers, and audit tag coverage
• /platform-skills:kubescape — score manifests, Helm charts, Kustomize or a live cluster against NSA, MITRE, CIS (EKS, AKS, GKE, vanilla) or SOC2, add a CI gate that fails PRs only on new findings, write exceptions that actually suppress, preview automatic fixes, and enforce controls with Kubescape's CEL admission policy library
• /platform-skills:kubescape-operator — install the Kubescape operator with Flux after a kernel and overlap preflight, rank CVEs by whether they load, alert on runtime behaviour against learned container profiles, turn generated NetworkPolicies into reviewed Git changes, accept risks with expiring SecurityExceptions, export metrics, and debug node-agent and scan failures
• /platform-skills:self-improve — bootstrap global or project-local .learnings/ workspace (init global/init local), log/recall/review/promote/revoke learnings and errors with provenance, scope and expiry, status overview, and migrate between scopes
• /platform-skills:supply-chain — sign images, generate and attest SBOMs, run CVE severity gates, enforce image signatures in Kubernetes, and generate SLSA Level 2 provenance
• /platform-skills:runtime-security — deploy Falco with eBPF, write custom rules, route alerts, debug why a rule is not firing, and bridge Falco signals to Kyverno admission enforcement
• /platform-skills:chaos — install Litmus Chaos or Chaos Mesh, generate fault experiments, schedule recurring chaos, run structured GameDay, debug stuck experiments, report results
• /platform-skills:dora — instrument DORA metrics in GitHub Actions, generate Grafana dashboards, benchmark against performance bands, debug missing metric data
• /platform-skills:awesome-docs — generate any animated Markdown document (README, architecture guide, runbook, tutorial, RFC, post-mortem, or custom), convert existing Markdown to animated, update diagrams, diff for staleness, audit quality, preview locally, or export to Confluence/Notion HTML
• /platform-skills:aws — CloudFront distributions, WAF web ACLs, Lambda@Edge, CloudFront Functions, Firewall Manager multi-account enforcement, and Terraform module generation with best practices
• /platform-skills:composite-actions — generate a full composite action repo scaffold, review an existing action.yml, harden with SHA pinning and env isolation, or generate a test workflow
• /platform-skills:github-actions — design reusable workflows and job graphs, review for OIDC/SHA pinning/token scoping, or debug failing CI workflows
• /platform-skills:zizmor — audit workflows, composite actions, Dependabot and pre-commit configs with zizmor: interactive three-layer wizard routes intent → scope → trade-offs (including an explicit hash-pin vs ref-pin decision, SHA recommended), then scan, deep-audit, auto-fix, triage findings, generate .github/zizmor.yml policy, emit a SARIF or hard-fail CI gate, or add the pre-commit hook; complements actionlint (/platform-skills:github-actions) rather than replacing it
• /platform-skills:kingfisher — find, live-validate, map the blast radius of, and revoke leaked secrets with Kingfisher across a repo, Git history, a GitHub/GitLab/Bitbucket org, S3/GCS, Docker images, Slack, Jira, Confluence, Teams, or Postman; scan, deep-audit, direct validate/revoke, baseline management, triage, kingfisher.yaml policy, CI gate, or pre-commit hook; complements the pattern-only secret scan in /platform-skills:trivy
• /platform-skills:kubernetes — generate cluster baseline scaffolds, diagnose 401/403 RBAC errors, harden Deployment specs, or debug crashloop/OOMKill/pending pods
• /platform-skills:azure — Workload Identity and OIDC federation, resource tagging with Azure Policy, AKS provisioning, RBAC scoping, and production-readiness review
• /platform-skills:openshift — SCC diagnosis and minimum grants, Route TLS modes and 503 debug, OpenShift GitOps app delivery, and cluster upgrade validation
• /platform-skills:secrets — design ESO vs Sealed Secrets strategy, scaffold SecretStore/ExternalSecret, seal and rotate secrets, run rotation runbooks, and audit Kubernetes-side secrets hygiene
• /platform-skills:renovate — generate renovate.json for any repo, or emit a GHA workflow to validate it on PR
• /platform-skills:setup-agents — scaffold multi-agent AI configs for any repo: ranked scan, interview-driven, generate/upgrade/add/review
• /platform-skills:ai-governance — generate real-time session hooks (Copilot, Claude Code) and a merge-time backstop that gate AI-agent edits to protected paths and dangerous commands; check dry-runs policy changes, audit reports fleet-wide adoption via gh api, explain translates an existing policy to plain English
• /platform-skills:token-optimizer — route broad repository discovery to a cheaper worker model using each client's native subagents; inspect detects clients and conflicts, setup writes a reviewable diff, doctor reports delegation, resolved model, redirection, and read limit as four separate states, explain dry-runs a classification, benchmark and report record and aggregate what the client exposes (reconciliation is against the provider's billing)
• Working Flux CD examples: examples/fluxcd/