49 Specialized Agents: A Complete Guide to AI-Powered Development

Why Claude Code Specialized Agents Matter

Ask a general-purpose AI to write a React component and you will get something that works. Ask it to write a React component with proper suspense boundaries, server component data fetching, and accessible keyboard navigation -- and the output starts to fall apart. The model has breadth but lacks the concentrated depth you need for production code.

One Shot Ship solves this by maintaining 49 specialized agents, each one a deep expert in a single domain. When you run /oss:build, the system reads your implementation plan, looks at each task, and dispatches it to the agent with the right expertise. A TypeScript type puzzle goes to typescript-pro. A Dockerfile goes to deployment-engineer. A slow SQL query goes to database-optimizer.

You do not need to know which agent to call. The routing is automatic. You focus on what to build; the agents focus on how to build it right.

The Full Agent Roster

Every agent below ships with One Shot Ship. They are organized by domain so you can see the coverage at a glance.

How Automatic Dispatch Works

When you run /oss:build, the build command reads your implementation plan -- the structured task list created by /oss:plan. Each task includes context about the technology involved: the file paths, the frameworks in use, the type of work (component, migration, test, deploy). The build command uses that context to select the best agent for the job.

The dispatch logic follows a simple principle: the agent with the deepest knowledge of the relevant domain handles the task. If a task touches both React and TypeScript, the system picks the agent whose expertise is most relevant to the primary concern. A component rendering issue goes to react-specialist. A complex generic type goes to typescript-pro.

Every agent enforces the same quality standards -- London-style TDD, no any types, no skipped tests -- but applies domain-specific best practices on top. A security-auditor checks for OWASP Top 10 vulnerabilities. A performance-engineer profiles bundle size and runtime bottlenecks. A database-optimizer reviews query plans and indexing strategies.

Example: One Feature, Three Agents

Suppose you are building a user settings page. The plan includes three tasks:

1. React component -- a settings form with tabs, validation, and optimistic updates. Dispatched to react-specialist, which produces a component using proper form state management, accessible tab navigation, and error boundaries.
2. API endpoint -- a PATCH route with input validation, rate limiting, and audit logging. Dispatched to backend-architect, which structures the route with Zod validation, proper HTTP status codes, and middleware composition.
3. Database migration -- adding a preferences JSONB column with a GIN index. Dispatched to database-optimizer, which writes the migration with proper indexing, a rollback path, and a data backfill strategy for existing rows.

All three tasks run under TDD. Each agent writes failing tests first, implements minimal code to pass, and refactors. The result is a feature where every layer was built by a domain expert, and every layer has full test coverage.

Beyond Code Generation

Agent specialization is not just about writing code. The security-auditor reviews authentication flows and dependency trees. The accessibility-specialist checks WCAG compliance and screen reader behavior. The incident-responder generates runbooks and post-mortem templates when production issues arise.

This is the difference between AI that writes code and AI that ships software. Shipping requires expertise across the entire lifecycle: planning, building, testing, deploying, monitoring, and responding to incidents. One Shot Ship covers every stage with a dedicated specialist.