ops-devops-platformDesigns DevOps and platform engineering systems. Use when planning Kubernetes, Terraform, GitOps, CI/CD, observability, incident response, or cloud-native operations.
Independent GitHub snapshot · Automated classification · Original documentation governs
| Need | Starting Direction |
|---|---|
| infrastructure provisioning | Terraform, OpenTofu, Pulumi, or cloud-native IaC |
| cluster or app deployment | GitOps first for steady-state, direct tooling for local iteration |
| CI/CD | protected pipelines plus workload identity and supply-chain controls — see supply-chain-security |
| observability | OpenTelemetry plus metrics, logs, traces, and SLO-based alerting |
| platform engineering | golden paths, policy-as-code, and self-service interfaces |
| incident operations | runbooks, severity model, escalation, and postmortems |
| Situation | Rule |
|---|---|
| infrastructure change | IaC by default; reconcile an emergency manual change back into code before the next promotion |
| steady-state production reconciliation | GitOps (Argo CD / Flux) over push-based deploys |
| CI credentials | workload identity (OIDC) over long-lived secrets |
| alerting | page on actionable SLO burn rates or imminent capacity exhaustion; send diagnostic host metrics to dashboards |
| new environments | platform template + policy guard; no snowflakes |
| supply-chain integrity | SLSA build track + cosign keyless signing |
| drift | detect via reconciler or terraform plan in CI; never discover by accident |
| Domain | Do | Anti-pattern to avoid |
|---|---|---|
| Provisioning | all material changes in IaC; explicit promotion gates | clickops drift; untagged infrastructure |
| Delivery | protected pipelines; artifact provenance; rollback + smoke checks | pipelines without identity boundaries |
| Platform | golden paths before self-service; policy-as-code that reduces variation | tools shipped without adoption path or ownership |
| Observability | define SLOs first; join logs/traces/metrics on shared trace ID | alert fatigue from raw host-metric thresholds |
| Incidents | postmortems feed runbooks and platform changes | postmortems that stop at narrative |
| Cost | tagging + budget alerts at resource creation; monthly right-sizing | unmanaged snowflake environments; unreviewed reservations |
| Load when… | Reference |
|---|---|
| supply-chain, SBOM, signing, SLSA | references/supply-chain-security.md |
| DORA's five metrics and team archetypes (Elite/High/Medium/Low tiers are retired), AI-adoption instability tax, general DevOps best practices | references/devops-best-practices.md |
| GitLab CI — parent/child pipelines, MR variable traps, env-export pattern | references/gitlab-ci-patterns.md |
| choosing a tool (IaC, GitOps, CI, policy, observability) | references/tool-landscape.md |
| golden paths, internal developer portal, platform maturity, when NOT to build an IDP, platform-vs-product boundary, measuring team cognitive load (Weis four-cluster model, Teamperature, leadership load), CI/IaC/GitOps adoption sequencing | references/platform-engineering-patterns.md |
| GitOps multi-env promotion, Argo CD / Flux patterns, automation lag and why continuous apply beats apply-on-change | references/gitops-workflows.md |
Terraform state isolation, why terraform workspace is wrong for environments, stage/prod/mgmt/global layout, secrets-in-state and backend choice | references/terraform-state-architecture.md |
| stack sizing (monolithic → application-group → service → micro), blast radius, "is my stack a monolith?" | references/stack-sizing-patterns.md |
| IaC testing rungs and their blind spots, infrastructure test diamond vs pyramid, Swiss-cheese layering | references/infrastructure-testing-strategy.md |
| on-call, severity model, escalation, postmortems | references/sre-incident-management.md |
| day-2 operational runbooks, environment hygiene | references/operational-patterns.md |
| AIOps alert correlation, automated triage | references/aiops-patterns.md |
| Kalman canary, cost autoscaler, CI capacity stabiliser | references/control-theory-applied.md |
| capacity planning, saturation SLO, pipeline bottleneck hunt | references/queueing-theory-applied.md |
| CI/CD throughput recovery, constraint surfacing, spend reallocation | references/theory-of-constraints-applied.md |
| platform-team charter, algedonic escalation, PRR audit | references/cybernetics-vsm-applied.md |
| MTBF/MTTR, availability budgets, FMEA | references/reliability-theory-applied.md |
| CAP/PACELC, consensus, idempotency, quorums | references/distributed-systems-applied.md |
| source URLs and release trackers | data/sources.json |
When changing the source inventory, run python3 scripts/validate_sources.py --skip-network from this skill directory for schema and coverage only. Run the offline CLI regressions with python3 -m pytest scripts/test_validate_sources.py. URL reachability and source content require separate checks; a schema pass does not refresh evidence dates.
AWS / GCP / Azure
Kubernetes
Docker / Kafka
Terraform / IaC
CI/CD and GitOps
Monitoring / Observability
Incident response
Security / Cost
When prior decisions or pitfalls are relevant, consult learnings.consolidated.md if present; use learnings.md only for needed history or as the available fallback. Otherwise skip both.
After applying it, if you encountered a pattern worth remembering, a mistake worth preventing, or a domain fact that surprised you, append one dated bullet to learnings.md via agents-skills-feedback-loop/scripts/append_learning.py. Do not modify SKILL.md itself.