Deployment Engineer

Provides guidance for shipping changes safely with pipelines, releases, progressive delivery, and operational guardrails.
Produces deployment workflow recommendations without requiring other skills.

Use this skill when

• Designing or improving CI/CD pipelines and release workflows
• Adding rollout safety (canary/blue-green), automated rollbacks, and runbooks
• Adding config validation gates and environment drift checks
• Implementing GitOps patterns (ArgoCD/Flux) at the workflow level

Do not use this skill when

• Cloud platform architecture (landing zones, network/IAM design)
• Deep Terraform/OpenTofu module design, state strategy, provider internals
• Deep FinOps cost optimization programs
• You only need to select cloud resources/services without deployment workflow changes

Required inputs

• Repo/tooling context (CI system, deployment tooling, runtime platform)
• Environments and promotion path
• Release cadence and change windows
• SLOs or error budget constraints
• Constraints (compliance, approvals, security requirements)

Workflow (Deterministic)

1. Capture inputs (repo/tooling, environments, release cadence, constraints, SLOs).
   • Output: input checklist and missing info questions.
   • If inputs are missing, stop and list questions before proceeding.
2. Map environments and promotion path (dev -> staging -> prod).
   • Output: environment map with promotion rules.
3. Define quality gates and config validation.
   • Output: gate list with owners, signals, and fail criteria.
4. Choose rollout + rollback strategy.
   • If traffic shaping exists, prefer canary; otherwise prefer rolling/blue-green.
   • Output: rollout plan, rollback triggers, and stop conditions.
5. Design pipeline stages and approvals.
   • Output: stage diagram or ordered list with required artifacts.
6. Add observability hooks and runbook steps.
   • Output: deploy markers/metrics list and runbook checklist.
7. Validate in staging and document operational handoff.
   • Output: validation checklist and open risks.

Common pitfalls to avoid

• Shipping without explicit rollback triggers or owners
• Allowing config drift between environments without checks
• Putting slow, flaky tests in early gates
• Using manual approvals without objective criteria
• Missing runbook steps for partial failures

Examples

Example request
"We need a GitHub Actions pipeline with canary deploys and automatic rollback. Add config validation before deploy."

Example response outline

• Pipeline: PR checks -> build -> test -> validate-config -> deploy-staging -> canary-prod -> verify
• Rollout: 10% canary for a defined window; rollback on 5xx > threshold
• Config validation: schema + env diff checks in validate-config
• Runbook: deploy, pause, rollback, and troubleshooting steps

Output format

Provide these sections in order:

1. Summary
2. Pipeline Stages & Gates
3. Rollout & Rollback Plan
4. Config Validation Strategy
5. Runbook & Observability
6. Open Questions / Risks

Output contract (Always)

• Pipeline stage diagram (or bullet list) with gates and required artifacts
• Rollout/rollback plan and stop conditions
• Config validation strategy (what is validated and where)
• Runbook notes (how to deploy, rollback, and troubleshoot)
• Open questions when required inputs are missing

References (Optional)

• references/README.md