Introduction
The Problem
AI coding assistants are generalists. They write code, but they don't enforce architecture, they don't follow your team's patterns, and they don't catch what a senior engineer would.
You end up:
- Re-explaining the same conventions every session
- Getting "it works" code that doesn't match your architecture
- Missing security issues, performance problems, or compliance gaps
- Doing manual code review on AI-generated code
The Solution
Specialist Agent gives your AI assistant a team of domain experts. Instead of one generalist, you get:
- @builder that knows your framework's component patterns
- @reviewer that enforces SOLID, security, and architecture rules
- @security that catches OWASP Top 10 issues
- @sentry-triage that triages production errors automatically
Each agent carries specific knowledge, rules, and workflows. They don't just write code — they enforce standards.
How It Works
You install agents into your project. When you need something done, you call the right agent:
> Use @builder to create the products module with CRUD
> Use @sentry-triage to check errors from the last 24h
> Use @reviewer to review src/modules/auth/
> /plan add real-time notifications with WebSocketAgent Categories
| Category | Agents | Purpose |
|---|---|---|
| Core | 5 | Build, review, debug, migrate |
| Workflow | 9 | Plan, execute, test, pair program |
| Engineering | 17 | API, security, data, DevOps, architecture |
| Business | 3 | Marketing, product, support |
| Automation | 2 | Sentry triage, autonomous builds |
Skills
Skills are repeatable workflows you invoke with a slash command:
| Category | Skills |
|---|---|
| Planning | /brainstorm, /plan, /discovery, /estimate |
| Development | /tdd, /debug, /autofix, /lint, /commit |
| Quality | /verify, /codereview, /audit, /health |
| Workflow | /checkpoint, /finish, /worktree, /learn |
| Migration | /migrate-framework, /migrate-architecture |
| Knowledge | /remember, /recall, /onboard, /tutorial, /write-skill |
What Makes It Different
| Feature | Generic AI | Specialist Agent |
|---|---|---|
| Architecture enforcement | No | Agents enforce your ARCHITECTURE.md |
| Framework-specific patterns | Generic | 7 framework packs with optimized agents |
| Code review depth | Surface-level | 3-in-1 review (quality + security + architecture) |
| Production ops | Manual | @sentry-triage auto-triages errors |
| Business agents | No | @marketing, @product, @support |
| Repeatable workflows | Manual | 32 skills (/plan, /tdd, /autopilot, ...) |
| Multi-platform | Varies | 6 platforms supported |
Framework Packs
Each framework gets optimized agents with framework-specific patterns:
Vue 3 | React | Next.js | SvelteKit | Angular | Astro | Nuxt
Platform Support
Works on: Claude Code, Cursor, VS Code, Windsurf, Codex, OpenCode
Who It's For
- Solo developers who want senior-level review on every commit
- Teams who want to standardize how AI assists development
- Startups who need to move fast without accumulating tech debt
- Enterprise teams who need compliance, security, and architecture enforcement
What's Next
- Quick Start — Install and use your first agent
- Agent Catalog — Browse all 36 agents interactively
- Skills Reference — All 32 skills