18 Claude Code subagent definitions forming a complete virtual engineering team
Agent
Use when selecting, benchmarking, or tuning Ollama models on constrained edge hardware; analysing LLM inference performance and context window behaviour; designing or evaluating prompt injection detection approaches; or producing model research documentation. Produces research notes, configuration recommendations, and…
Agent
Use when building or modifying server-side services, REST or GraphQL APIs, background workers, CouchDB data models, or Go microservices. Covers Node.js/TypeScript and Go implementations.
Agent
Use when capturing or refining requirements, writing user stories, defining acceptance criteria, producing BDD Gherkin scenarios, mapping user journeys, or modelling domain concepts. The upstream anchor for all feature work — nothing enters the backlog without BA sign-off.
Agent
Use when reviewing a pull request or a code change for correctness, security, performance, maintainability, and adherence to team standards. Reads only — produces a structured review report.
Agent
Use when you need to plan, draft, or revise written content across three modes: general content and communications (announcements, blog posts, newsletters, social copy, talking points), technical documentation (guides, references, tutorials, API docs), or release notes (grouped, user-facing summaries of shipped…
Agent
Use immediately before promoting a build to staging or production. Validates environment configuration, migration readiness, rollback plan, monitoring baseline, and SLO health. Reads only — produces a signed-off checklist artifact.
Agent
Use when designing or modifying CI/CD pipelines, GitHub Actions workflows, OpenTofu infrastructure, Docker images, ECS task definitions, ECR configuration, or deployment strategies. Produces configuration and IaC artifacts and can execute deployments via Bash.
Agent
Use when building or modifying UI components, pages, client-side state, web performance, accessibility, or browser-facing TypeScript. Covers React/Next.js with TypeScript.
Agent
Use when delivering a self-contained feature that spans database, API, and UI layers together and splitting the work across backend-developer and frontend-developer would create unnecessary coordination overhead.
Agent
Use when working on bare-metal or edge Linux configuration — systemd service units and drop-ins, kernel sysctl hardening, SSH configuration, UFW firewall rules, fail2ban, apt/unattended-upgrades, cgroup v2, boot configuration, or ARM64 Linux deployment. Produces configuration files and documentation artifacts, and can…
Agent
Use when designing or evolving the internal developer platform: Backstage service catalog, golden path templates, developer self-service workflows, ArgoCD GitOps configuration, Kong API gateway routing, or Linkerd service mesh policy.
Agent
Use when writing or modifying Python code. Covers Python 3.9+ best practices, stdlib-first development, systemd-compatible logging, environment-variable-driven configuration, and test-driven development for edge services, CLI tools, HTTP servers, and automation.
Agent
Use when designing test strategies, writing or reviewing test suites, setting coverage standards, configuring Playwright E2E tests, designing k6 load test scenarios, or assessing test pyramid health.
Agent
Use when facilitating sprint ceremonies (as written artifacts), writing Definition of Done, maintaining the impediment log, producing retrospective documents, reviewing sprint health, or assessing team process against Scrum or Kanban standards.
Agent
Use when conducting security reviews, threat modelling, defining security controls, reviewing dependency vulnerabilities, configuring secrets management, designing IAM policies, assessing DevSecOps pipeline security, or reviewing infrastructure security on bare-metal and edge hosts. Produces configuration and policy…
Agent
Use when defining SLOs, error budgets, or SLIs; designing alerting and on-call runbooks; reviewing system reliability and toil; configuring Prometheus recording rules and Grafana dashboards; or conducting post-incident reviews. Can execute operational checks and diagnostics via Bash.
Agent
Use when facing a bug, unexpected behaviour, production incident, or test failure that is not immediately obvious. Applies a structured hypothesis-driven methodology to find root cause. Reads code and logs — does not modify files.
Agent
Use when making or reviewing significant design decisions: new services, API contracts, data modelling, cross-cutting concerns, technology selection, or any change that affects system boundaries or quality attributes. Produces Architecture Decision Records (ADRs) and C4 diagrams in Markdown.
Agent
Use when defining or evolving the design system, design tokens, typography scale, colour palette, component specifications, or design-to-code handoff. Produces Tailwind configuration, design token JSON, component specs, and accessibility-led visual standards. Does not produce image or graphic assets.