11 Specialized Agents (Consolidated in v1.17.0)
Agents auto-activate based on your prompt context. v1.17.0 consolidated 15 agents into 11 for better routing:
Development (4)
| Agent | Specialization |
|-------|---------------|
| architect | System design, database, backend (Node.js, Python, Laravel, Go) β NEW: replaces backend-expert + database-specialist |
| ui-expert | Frontend (React, Vue, Angular, Next.js) + design systems β NEW: replaces web-expert + ui-designer |
| mobile-expert | React Native, Flutter β Expo, cross-platform, NativeWind |
| game-developer | Godot game development, multi-platform export |
Quality & Security (2)
| Agent | Specialization |
|-------|---------------|
| security-expert | OWASP audits, vulnerability scanning |
| qa-automation | Jest, Cypress, Detox, testing strategies |
Infrastructure (2)
| Agent | Specialization |
|-------|---------------|
| devops-cicd | Docker, K8s, CI/CD, monitoring |
| voice-operations | ElevenLabs AI narration |
System (3)
| Agent | Specialization |
|-------|---------------|
| project-manager | Project detection, config loading, context β NEW: replaces 3 agents |
| smart-agent-detector | Intelligent agent + model selection |
| pm-operations-orchestrator | Workflow coordination |
---
6 MCP Servers
MCP (Model Context Protocol) servers auto-invoke based on context:
| MCP Server | Purpose | Auto-Triggers | Setup |
|------------|---------|---------------|-------|
| context7 | Library documentation | "Build with MUI", library names | None |
| playwright | Browser automation, E2E | "Test the login page" | None |
| vitest | Test execution, coverage | "Run tests", "Check coverage" | None |
| firebase | Firebase services | "Set up Firestore", "Firebase Auth" | firebase login |
| figma | Design file fetching | Figma URLs | FIGMA_API_TOKEN |
| slack | Notifications | Phase 9 completion | SLACK_BOT_TOKEN |
MCPs requiring tokens will silently skip if not configured.
```bash
# Context7 auto-fetches React docs
"Build a form with Material UI"
# Firebase manages your project
"Set up Firestore for my app"
```
See: [aura-frog/docs/MCP_GUIDE.md](aura-frog/docs/MCP_GUIDE.md) for setup
---
Skills System (Optimized in v1.17.0)
Skills activate automatically based on your message context. v1.17.0 introduced skill bundles that lazy-load patterns on-demand:
```
User: "Implement user profile screen"
β
Auto-invokes:
1. agent-detector β Selects mobile-expert + model
2. model-router β Picks Sonnet (standard task)
3. project-context-loader β Loads your conventions
4. workflow-orchestrator β Executes 9-phase workflow
```
13 Auto-Invoking Skills:
| Skill | Triggers | Purpose |
|-------|----------|---------|
| agent-detector | Every message | Select agent + complexity |
| model-router | After agent detection | Select Haiku/Sonnet/Opus |
| workflow-orchestrator | "implement", "build", "create" | 9-phase workflow |
| project-context-loader | Before code generation | Load conventions |
| framework-expert | Framework tasks | Lazy-load React/Vue/etc patterns |
| seo-bundle | SEO/GEO tasks | Lazy-load SEO patterns |
| testing-patterns | Test tasks | Universal testing patterns |
| bugfix-quick | "fix", "error", "broken" | Fast TDD bug fixes |
| test-writer | "add tests", "coverage" | Generate tests |
| code-reviewer