Agents Reference
Complete reference for all Mannay agents (domain specialists).
Overview
Agents are domain specialists invoked by skills or directly. They provide expert knowledge in specific areas.
How Agents Are Invoked
- By skills automatically - Skills invoke relevant agents based on task
- By the Task tool -
Task(subagent_type="agent-name") - By mentioning their domain - Orchestrator detects and invokes
Architecture Agents
backend-architect
Expertise: API design, database schema, authentication, reliability
Invoke when:
- Designing APIs or database schemas
- Implementing authentication
- Building reliable backend systems
Standards:
- Data integrity first
- Fault tolerance
- Security by default
frontend-architect
Expertise: Component architecture, responsive design, Core Web Vitals
Invoke when:
- Designing component hierarchies
- Optimizing frontend performance
- Building design systems
Standards:
- Lighthouse ≥95
- Mobile-first
- Accessibility built-in
api-designer
Expertise: REST, GraphQL, tRPC, gRPC, contract-first design
Invoke when:
- Designing API contracts
- Creating OpenAPI specs
- Building GraphQL schemas
Standards:
- Contract-first
- Versioning strategy
- Clear error responses
system-architect
Expertise: Scalability, component boundaries, long-term decisions
Invoke when:
- Planning system architecture
- Making technology decisions
- Designing for scale (10x growth)
Standards:
- Maintainability focus
- Clear boundaries
- Documentation
requirements-analyst
Expertise: PRD creation, user stories, scope definition
Invoke when:
- Gathering requirements
- Writing user stories
- Defining scope
Standards:
- Structured discovery
- Testable acceptance criteria
- Clear scope boundaries
tech-stack-researcher
Expertise: Technology comparisons, library recommendations
Invoke when:
- Choosing technologies
- Comparing libraries
- Evaluating trade-offs
Standards:
- Objective comparison
- Cost consideration
- Future-proofing
Quality Agents
code-reviewer
Expertise: SOLID principles, code quality, production readiness
Invoke when:
- Reviewing code before merge
- Checking quality standards
- Finding improvements
Standards:
- Quantified metrics
- Actionable feedback
- Production focus
typescript-pro
Expertise: Advanced types, type safety, strict mode
Invoke when:
- Complex type definitions
- Type inference issues
- Generic patterns
Standards:
- No
anytypes - Full type coverage
- Strict mode compliance
accessibility-specialist
Expertise: WCAG 2.1 AA, screen readers, keyboard navigation
Invoke when:
- Building UI components
- Auditing accessibility
- Form design
Standards:
- WCAG 2.1 AA minimum
- Keyboard navigable
- Screen reader tested
performance-engineer
Expertise: Bottleneck identification, optimization, profiling
Invoke when:
- Performance issues
- Optimization needed
- Bundle size concerns
Standards:
- Measurement-driven
- Lighthouse ≥95
- Bundle <50KB gzipped
security-engineer
Expertise: Vulnerability assessment, OWASP Top 10, zero-trust
Invoke when:
- Authentication/authorization
- Input validation
- Security review
Standards:
- OWASP compliance
- Input validation
- Security-first design
refactoring-expert
Expertise: Technical debt reduction, clean code, DRY
Invoke when:
- Code cleanup needed
- Technical debt
- Improving maintainability
Standards:
- Clean code principles
- SOLID adherence
- Test coverage maintained
Documentation Agents
technical-writer
Expertise: Clear documentation, audience-tailored content
Invoke when:
- Writing documentation
- API documentation
- User guides
Standards:
- Clear and concise
- Audience appropriate
- Examples included
documentation-engineer
Expertise: Docs-as-infrastructure, automated generation
Invoke when:
- Setting up doc systems
- Automated doc generation
- Multi-format output
Standards:
- Versioned docs
- Automated testing
- CI/CD integration
learning-guide
Expertise: Progressive teaching, concept explanation
Invoke when:
- Explaining concepts
- Teaching patterns
- Learning materials
Standards:
- Progressive complexity
- Practical examples
- Clear explanations
deep-research-agent
Expertise: Comprehensive research, analysis
Invoke when:
- Deep investigation needed
- Multi-source research
- Complex analysis
Standards:
- Thorough coverage
- Source verification
- Structured output
Agent Chaining
Skills automatically chain agents based on task domain:
| Task Domain | Agents Chained |
|---|---|
| User authentication | security-engineer → backend-architect → api-designer |
| UI component | frontend-architect → accessibility-specialist → typescript-pro |
| API endpoint | api-designer → backend-architect → security-engineer |
| Database work | backend-architect → typescript-pro |
| Performance issue | performance-engineer → frontend-architect or backend-architect |