deployment-engineer · qdhenry — Claude Code subagent: CI/CD and Kubernetes
deployment-engineer — a Claude subagent that drafts CI/CD pipelines, container builds and cloud infrastructure, then hands over a deployment runbook.
no license ★ 1,357 updated 2026-03-01
.claude/agents/deployment-engineer.mdWhat deployment-engineer does
The subagent runs on sonnet and can edit files, search the repo and run shell commands. What comes back: a commented pipeline file for GitHub Actions, GitLab CI or Jenkins, a multi-stage Dockerfile, Kubernetes manifests or a Helm chart, Terraform or CloudFormation samples.
Its working order is fixed. Read the existing patterns first, then write code, with tests kept alongside. A build that fails lint, type checks, security scans or tests is not finished, so nothing merges on red. Infrastructure lives in code and Git stays the source of truth: changes arrive as pull requests and get reconciled into the environment. Faced with several valid designs, it ranks them by testability, readability, fit with current patterns, simplicity and how easily the choice can be reversed.
Two MCP servers show up in its tool list, mcp__context7__* and mcp__sequential-thinking__*. Connect context7 and sequential-thinking, or those calls go nowhere and the agent falls back on what it already knows. Installation is one file at .claude/agents/deployment-engineer.md.
What it helps with
- write a GitHub Actions pipeline with lint, tests and image scanning
- shrink a Docker image and stop it running as root
- generate Kubernetes manifests for a deployment, service and ingress
- provision cloud resources with Terraform
- document manual rollback steps in
RUNBOOK.md
Example prompts for the agent
- Set up a pipeline for this repo: lint, tests, build the image, deploy to staging.
- Turn our Kubernetes manifests into a Helm chart and explain the structure you picked.
- Write the manual rollback steps into
RUNBOOK.mdin case the automated one fails.
How to install deployment-engineer
Manually: put the files in place
.claude/agents/deployment-engineer.mdFor all projects use ~/.claude/agents/. Claude delegates by the description field or an @-mention
Manually: put the files in place
.codex/agents/deployment-engineer.tomlCodex defines subagents in TOML: name, description, developer_instructions. Move the Markdown text into developer_instructions by hand. For all projects use ~/.codex/agents/
Manually: put the files in place
.cursor/agents/deployment-engineer.mdCursor also reads .claude/agents/
Step by step, with commands for each platform: Claude Code subagents: how to create one, call it and keep context clean
Questions about deployment-engineer
How do I install deployment-engineer in Claude Code?
Put the subagent file at .claude/agents/deployment-engineer.md inside your project; Claude reads subagents from that folder.
What does deployment-engineer do?
It produces pipeline configuration, an optimized Dockerfile, Kubernetes manifests or a Helm chart, and infrastructure-as-code samples. It also works out how per-environment configs get injected and writes a RUNBOOK.md covering rollback.
Does it need any MCP servers?
Yes. Its tools include context7 for library and cloud service docs, and sequential-thinking for working through infrastructure decisions.
Similar subagents
All subagents for DevOps and infrastructure →deployment-engineer by other authors comes first.