Markdown

SDD Agents

<!-- This file is maintained by the docs-writer agent. To update, invoke the docs-writer agent with your changes. -->

The specialized agents that power SDD.

How Agents Work

SDD uses specialized agents instead of one general-purpose AI. Each agent has:

  • A specific area of expertise
  • Tools appropriate to its role
  • Standards and patterns it follows

You can invoke agents in two ways:

  1. **Via `/sdd`** - Commands like `/sdd I want to start implementing` automatically orchestrate agents based on the implementation plan
  2. **Directly** - Ask Claude to use an agent (e.g., "Use the backend-dev agent to implement this endpoint")

The Agents

api-designer

Designs API contracts.

**When it's used:** During implementation when API changes are needed **What it does:** Writes OpenAPI specs, generates TypeScript types for server and client

---

backend-dev

Implements server-side code.

**When it's used:** During implementation for backend work **What it does:** Writes Node.js/TypeScript code following CMDO architecture, writes unit tests

---

frontend-dev

Implements client-side code.

**When it's used:** During implementation for frontend work **What it does:** Writes React components following MVVM patterns, writes component tests

---

db-advisor

Reviews database design.

**When it's used:** When database changes are proposed **What it does:** Reviews schema and queries for performance, suggests optimizations

---

tester

Writes integration and E2E tests.

**When it's used:** During implementation for non-unit tests **What it does:** Creates Testkube test definitions, writes E2E scenarios

---

devops

Handles infrastructure and CI/CD pipelines.

**When it's used:** When deployment configuration or build automation is needed **What it does:** Writes Helm charts, Kubernetes configs, container definitions, GitHub Actions workflows, PR checks

---

reviewer

Reviews code and specs.

**When it's used:** During verification (`/sdd I want to verify the implementation`) **What it does:** Checks that implementation matches spec, reviews code quality

---

Agent Models

Agents use different models based on their task complexity:

| Role | Model | Agents | |------|-------|--------| | Strategic | Opus | reviewer, db-advisor | | Implementation | Sonnet | api-designer, backend-dev, frontend-dev, tester, devops |

Strategic agents handle decisions that require deep analysis and affect the whole project. Implementation agents execute defined tasks quickly and efficiently.

Next Steps

  • [Getting Started](getting-started.md) - See agents in action
  • [Workflows](workflows.md) - How agents work together
  • [Commands](commands.md) - Commands that orchestrate agents