specialist-agent

Specialist Agent

Your AI development team. 36 agents. 32 skills. Ship faster.

npx specialist-agent init

"Use @builder to create a payments module with CRUD"
"Use @reviewer to review the auth module"
/tdd implement calculateDiscount function

Core Agents

The foundation. Build, review, debug, migrate.

Agent	What it does	Example
@starter	Scaffold new projects from scratch	"Create a Next.js + Prisma + Stripe app"
@builder	Build modules, components, services	"Create a products module with CRUD"
@reviewer	3-in-1 review: spec + quality + architecture	"Review the auth module"
@doctor	Systematic 4-phase debugging	"Investigate the 500 error on login"
@migrator	Modernize legacy code	"Migrate src/legacy/ to TypeScript"

Workflow Agents

Plan, execute, coordinate, remember.

Agent	What it does	Example
@planner	Adaptive planning by complexity	"Plan the checkout flow"
@executor	Execute plans with checkpoints and gates	"Execute the auth migration plan"
@tdd	Test-Driven Development (RED-GREEN-REFACTOR)	"TDD the discount calculator"
@debugger	Systematic debugging with hypotheses	"Debug the state sync issue"
@pair	Real-time pair programming	"Pair with me on the payment flow"
@analyst	Requirements to technical specs	"Spec out the notification system"
@orchestrator	Coordinate multiple agents on complex tasks	"Orchestrate the full checkout feature"
@scout	Quick project analysis	"Analyze this project"
@memory	Cross-session decision persistence	"Remember: use Zustand for state"

Engineering Agents

Domain specialists for every layer of your stack.

Agent	What it does	Example
@api	REST/GraphQL API design with OpenAPI	"Design the orders API"
@perf	Performance profiling and optimization	"Optimize the dashboard load time"
@i18n	Internationalization	"Add Portuguese support"
@docs	Documentation generation	"Generate API docs from code"
@refactor	Code refactoring with safety nets	"Extract service layer from controllers"
@deps	Dependency management and updates	"Audit and update dependencies"
@explorer	Deep codebase exploration	"Map the module dependencies"
@finance	Payments, billing, subscriptions	"Integrate Stripe checkout"
@cloud	AWS, GCP, Terraform, serverless	"Set up Lambda + API Gateway"
@security	Auth, OWASP Top 10, encryption	"Audit for XSS and injection"
@designer	Design systems, accessibility	"Create the component library"
@data	Database design, migrations, caching	"Design the schema with Prisma"
@devops	Docker, Kubernetes, CI/CD	"Create the GitHub Actions pipeline"
@tester	Test strategies and coverage	"Create test strategy for payments"
@legal	GDPR, LGPD, CCPA compliance	"Check LGPD compliance"
@architect	Architecture migration (DDD, Clean, Hexagonal)	"Migrate to Clean Architecture"
@ripple	Cascading effect analysis	"What breaks if I change the User model?"

Business Agents

Product, growth, and support.

Agent	What it does	Example
@marketing	Landing pages, copy, SEO, growth analytics	"Create the pricing page with SEO"
@product	Roadmaps, specs, PRDs, user stories	"Write PRD for the onboarding flow"
@support	Help docs, FAQs, runbooks, error pages	"Create the help center"

Automation Agents

Autonomous triage and resolution.

Agent	What it does	Example
@sentry-triage	Pull Sentry errors, cross-reference PRs, prioritize, auto-fix	"Triage today's Sentry errors"
@autopilot	Iterative autonomous builds with PRD and progress tracking	"Build the dashboard module overnight"

Skills

Slash commands that trigger repeatable workflows.

Planning

Skill	What it does
/brainstorm	Socratic brainstorming before planning
/plan	Adaptive feature planning
/discovery	Feasibility evaluation with GO/NO-GO verdict
/estimate	Estimate token cost before starting

Development

Skill	What it does
/tdd	Test-driven development cycle
/debug	4-phase systematic debugging
/autofix	Automatic error resolution
/lint	Lint and auto-fix (Biome/ESLint/Prettier)
/commit	Smart conventional commits with validation

Quality

Skill	What it does
/verify	Proof-based verification before completion
/codereview	Multi-reviewer parallel code review
/audit	Multi-domain audit: security + perf + arch + observability + deps
/health	Project health score (0-100)
/grill	Adversarial code challenge with attack vectors
/design-review	Frontend design audit (consistency, a11y, responsive, UX)
/seo-audit	SEO technical audit (meta, structured data, CWV)
/improve-architecture	Incremental architecture improvement

Workflow

Skill	What it does
/checkpoint	Save/restore progress with git
/finish	Finalize branch with metrics
/worktree	Git worktree isolation for parallel tasks
/learn	Learning mode, explains while building
/autopilot	Autonomous iterative development (PRD + progress + fresh context)

Migration

Skill	What it does
/migrate-framework	Migrate between frameworks
/migrate-architecture	Migrate architecture patterns (MVC, Clean, DDD, Hexagonal)

Product & Marketing

Skill	What it does
/prd	Product requirements with user stories and issues
/copywriting	Marketing copy with A/B variants
/cro	Conversion rate optimization audit

Knowledge

Skill	What it does
/remember	Save a decision to memory
/recall	Query saved decisions
/onboard	Codebase onboarding for new developers
/tutorial	Interactive guided tutorial
/write-skill	Create or improve custom skills

Tip: Use /btw (native Claude Code) for quick side questions without consuming tokens or polluting context. Perfect for mid-task clarifications while agents are working.

Framework Packs

7 packs, each with specialized agents, lite agents, and skills tailored to the framework's conventions.

Next.js | React | Vue 3 | SvelteKit | Angular | Astro | Nuxt

The installer auto-detects your framework. Each pack follows the idioms of its ecosystem: server components for Next.js, composables for Vue, stores for SvelteKit, signals for Angular, islands for Astro, auto-imports for Nuxt.

How It Works

Verification protocol every agent must show actual command output before claiming completion. No "should work." Run the command, show the output, then claim done.

Anti-rationalization built-in tables prevent agents from cutting corners or skipping steps.

Context isolation each subtask gets fresh context. No accumulated pollution across long sessions.

Full vs Lite every agent has a Lite variant. Full mode for complex work, Lite for scaffolding and quick fixes at 60-80% lower cost.

Platforms

Works with Claude Code, Cursor, VS Code, Windsurf, Codex, and OpenCode.

npx specialist-agent init          # Install agents + skills
npx specialist-agent detect        # Detect framework and architecture
npx specialist-agent list          # List installed agents

Documentation

specialistagent.com.br

Contributing

git checkout -b feature/my-feature
git commit -m 'feat: add my feature'
git push origin feature/my-feature

License

MIT