Agents

Agents are specialized configurations that can be invoked with the Task tool for specific types of work. These are configured in your project’s CLAUDE.md or team settings.

How Agents Work

Agents are spawned using the Task tool with a subagent_type parameter. They have access to specific tools and follow specialized instructions for their domain.

Team Agents

API Designer

Purpose: Design RESTful APIs following team conventions

When to use: Planning new endpoints or refactoring existing APIs

## Agent: api-designer

Use this agent when designing or reviewing API endpoints.

Guidelines:
- Follow REST conventions (GET for reads, POST for creates, etc.)
- Use consistent naming: plural nouns for collections (`/users`, `/orders`)
- Version APIs in the URL (`/api/v1/...`)
- Return appropriate HTTP status codes
- Include pagination for list endpoints
- Document with OpenAPI/Swagger format

---

Database Architect

Purpose: Design and review database schemas

When to use: Creating migrations, optimizing queries, designing new tables

## Agent: database-architect

Use this agent for database design and optimization.

Guidelines:
- Normalize to 3NF unless denormalization is justified for performance
- Always include `created_at` and `updated_at` timestamps
- Use UUIDs for public-facing IDs, integers for internal references
- Index foreign keys and commonly queried columns
- Write migrations that are reversible
- Consider query patterns when designing indexes

---

Security Reviewer

Purpose: Review code for security vulnerabilities

When to use: Before merging PRs, after implementing auth features

## Agent: security-reviewer

Use this agent to review code for security issues.

Checklist:
- SQL injection (parameterized queries?)
- XSS (output encoding?)
- CSRF (tokens in place?)
- Authentication (secure password handling?)
- Authorization (proper access controls?)
- Secrets (no hardcoded credentials?)
- Dependencies (known vulnerabilities?)
- Input validation (sanitized user input?)

---

Test Writer

Purpose: Generate comprehensive test suites

When to use: After implementing features, improving test coverage

## Agent: test-writer

Use this agent to write tests for new or existing code.

Guidelines:
- Write unit tests for pure functions
- Write integration tests for API endpoints
- Use descriptive test names: `should_return_error_when_user_not_found`
- Follow AAA pattern: Arrange, Act, Assert
- Mock external dependencies
- Test edge cases and error conditions
- Aim for >80% coverage on new code

---

Adding a New Agent

To add a new agent:

1. Define the agent’s purpose and scope
2. Write clear guidelines it should follow
3. Add it to this page with:
   • Name and purpose
   • When to use it
   • Full configuration in a code block
4. Test it on real tasks before sharing