Initial commit

This commit is contained in:
Zhongwei Li
2025-11-30 08:37:14 +08:00
commit 353f62980d
17 changed files with 4887 additions and 0 deletions

View File

@@ -0,0 +1,460 @@
---
name: using-dev-team
description: |
10 specialist developer agents for backend (Go/TypeScript/Python), DevOps, frontend,
design, QA, and SRE. Dispatch when you need deep technology expertise.
trigger: |
- Need deep expertise for specific technology (Go, TypeScript, Python)
- Building infrastructure/CI-CD → devops-engineer
- Frontend with design focus → frontend-designer
- Test strategy needed → qa-analyst
- Reliability/monitoring → sre
skip_when: |
- General code review → use default plugin reviewers
- Planning/design → use brainstorming
- Debugging → use systematic-debugging
related:
similar: [using-ring]
---
# Using Ring Developer Specialists
The ring-dev-team plugin provides 10 specialized developer agents. Use them via `Task tool with subagent_type:`.
**Remember:** Follow the **ORCHESTRATOR principle** from `using-ring`. Dispatch agents to handle complexity; don't operate tools directly.
---
## 10 Developer Specialists
### 1. Backend Engineer (Language-Agnostic)
**`ring-dev-team:backend-engineer`**
**Specializations:**
- Language-agnostic backend design (adapts to Go/TypeScript/Python/Java/Rust)
- Microservices architecture patterns
- API design (REST, GraphQL, gRPC)
- Database modeling & optimization
- Authentication & authorization
- Message queues & event-driven systems
- Caching strategies
**Use When:**
- You need backend expertise without language commitment
- Designing multi-language systems
- Comparing backend approaches across languages
- Architecture that might use multiple languages
- Language choice is not yet decided
**Example dispatch:**
```
Task tool:
subagent_type: "ring-dev-team:backend-engineer"
model: "opus"
prompt: "Design a user authentication service, recommend the best language and explain trade-offs"
```
---
### 2. Backend Engineer (Go)
**`ring-dev-team:backend-engineer-golang`**
**Specializations:**
- Go microservices & API design
- Database optimization (PostgreSQL, MongoDB)
- Message queues (Kafka, RabbitMQ)
- OAuth2, JWT, API security
- gRPC and performance optimization
**Use When:**
- Designing Go services from scratch
- Optimizing database queries
- Implementing authentication/authorization
- Troubleshooting concurrency issues
- Reviewing Go backend architecture
**Example dispatch:**
```
Task tool:
subagent_type: "ring-dev-team:backend-engineer-golang"
model: "opus"
prompt: "Design a Go service for user authentication with JWT and OAuth2 support"
```
---
### 3. Backend Engineer (TypeScript/Node.js)
**`ring-dev-team:backend-engineer-typescript`**
**Specializations:**
- TypeScript/Node.js backend services
- Express, Fastify, NestJS frameworks
- TypeScript type safety & ORM (Prisma, TypeORM)
- Node.js performance optimization
- Async/await patterns & error handling
- REST & GraphQL APIs
- Testing with Jest/Vitest
**Use When:**
- Building TypeScript/Node.js backends
- Migrating from JavaScript to TypeScript
- NestJS or Express service design
- Node.js-specific optimization
- Full-stack TypeScript projects
**Example dispatch:**
```
Task tool:
subagent_type: "ring-dev-team:backend-engineer-typescript"
model: "opus"
prompt: "Design a NestJS service with Prisma ORM for user management"
```
---
### 4. Backend Engineer (Python)
**`ring-dev-team:backend-engineer-python`**
**Specializations:**
- Python backend services (FastAPI, Django, Flask)
- SQLAlchemy & Django ORM
- Async Python (asyncio, aiohttp)
- Data processing & API integration
- Python type hints & validation (Pydantic)
- Celery for background tasks
- Testing with pytest
**Use When:**
- Building Python backend services
- FastAPI or Django projects
- Data-heavy backend services
- Machine learning API integration
- Python-specific optimization
**Example dispatch:**
```
Task tool:
subagent_type: "ring-dev-team:backend-engineer-python"
model: "opus"
prompt: "Design a FastAPI service with SQLAlchemy for e-commerce orders"
```
---
### 5. DevOps Engineer
**`ring-dev-team:devops-engineer`**
**Specializations:**
- CI/CD pipelines (GitHub Actions, GitLab CI)
- Containerization (Docker, Docker Compose)
- Kubernetes deployment & scaling
- Infrastructure as Code (Terraform, Helm)
- Cloud infrastructure setup
**Use When:**
- Setting up deployment pipelines
- Containerizing applications
- Managing Kubernetes clusters
- Infrastructure provisioning
- Automating infrastructure changes
**Example dispatch:**
```
Task tool:
subagent_type: "ring-dev-team:devops-engineer"
model: "opus"
prompt: "Create a GitHub Actions CI/CD pipeline for Go service deployment to Kubernetes"
```
---
### 6. Frontend Engineer
**`ring-dev-team:frontend-engineer`**
**Specializations:**
- React/Next.js application architecture
- TypeScript for type safety
- State management (Redux, Zustand, Context)
- Component design patterns
- Form handling and validation
- CSS-in-JS and styling solutions
**Use When:**
- Building React/Next.js applications
- Implementing complex UI components
- State management design
- Performance optimization for frontend
- Accessibility improvements
**Example dispatch:**
```
Task tool:
subagent_type: "ring-dev-team:frontend-engineer"
model: "opus"
prompt: "Design a React dashboard with real-time data updates and TypeScript"
```
---
### 7. Frontend Engineer (TypeScript)
**`ring-dev-team:frontend-engineer-typescript`**
**Specializations:**
- TypeScript-first React/Next.js architecture
- Advanced TypeScript patterns (generics, utilities, branded types)
- Type-safe state management (Zustand, Redux Toolkit)
- Type-safe API integration (tRPC, React Query)
- Component library with strict TypeScript
- Server Components with TypeScript (Next.js App Router)
- Type-safe routing & form validation
**Use When:**
- Building TypeScript-first frontend projects
- Requiring advanced type safety (e.g., branded types, strict inference)
- Type-safe full-stack integration (tRPC)
- Migrating JavaScript frontend to TypeScript
- Complex TypeScript patterns (conditional types, mapped types)
**Example dispatch:**
```
Task tool:
subagent_type: "ring-dev-team:frontend-engineer-typescript"
model: "opus"
prompt: "Design a type-safe React dashboard with tRPC, Zod validation, and branded types"
```
---
### 8. Frontend Designer
**`ring-dev-team:frontend-designer`**
**Specializations:**
- Distinctive, production-grade UI with high visual design quality
- Bold typography with characterful, non-generic fonts
- Cohesive color systems with dominant colors and sharp accents
- High-impact animations and micro-interactions
- Unexpected layouts with asymmetry, overlap, and spatial tension
- Atmosphere through textures, gradients, shadows, and visual depth
**Use When:**
- Building interfaces where aesthetics matter (landing pages, portfolios, marketing)
- Creating memorable, visually striking user experiences
- Designing dashboards with distinctive visual identity
- Implementing creative motion and interaction design
- Establishing distinctive visual languages and design systems
**Example dispatch:**
```
Task tool:
subagent_type: "ring-dev-team:frontend-designer"
model: "opus"
prompt: "Create a brutalist landing page for a tech startup with bold typography and unexpected layouts"
```
**Note:** Use frontend-designer for visual aesthetics and design excellence. Use frontend-engineer for complex state management, business logic, and application architecture.
---
### 9. QA Analyst
**`ring-dev-team:qa-analyst`**
**Specializations:**
- Test strategy & planning
- E2E test automation (Cypress, Playwright)
- Unit test coverage analysis
- API testing strategies
- Performance testing
- Compliance validation
**Use When:**
- Planning testing strategy for features
- Setting up E2E test suites
- Improving test coverage
- API testing and validation
- Quality gate design
**Example dispatch:**
```
Task tool:
subagent_type: "ring-dev-team:qa-analyst"
model: "opus"
prompt: "Create a comprehensive E2E test strategy for user authentication flow"
```
---
### 10. Site Reliability Engineer (SRE)
**`ring-dev-team:sre`**
**Specializations:**
- Observability (monitoring, logging, tracing)
- Alerting strategies & SLOs
- Incident response automation
- Performance optimization
- Scalability analysis
- System reliability patterns
**Use When:**
- Designing monitoring/observability systems
- Setting up alerts and SLOs
- Incident response planning
- Performance bottleneck analysis
- Reliability engineering
**Example dispatch:**
```
Task tool:
subagent_type: "ring-dev-team:sre"
model: "opus"
prompt: "Design monitoring and alerting for a Go microservice with 99.9% SLO"
```
---
## Decision Matrix: Which Specialist?
| Need | Specialist | Use Case |
|------|-----------|----------|
| Backend (language-agnostic, multi-language) | Backend Engineer | Architecture without language commitment |
| Go API, database, concurrency | Backend Engineer (Go) | Go-specific service architecture |
| TypeScript/Node.js backend, NestJS, Express | Backend Engineer (TypeScript) | TypeScript backend services |
| Python backend, FastAPI, Django, data | Backend Engineer (Python) | Python-specific backend services |
| CI/CD, Docker, Kubernetes, IaC | DevOps Engineer | Deployment pipelines, infrastructure |
| React, TypeScript, components, state | Frontend Engineer | General UI development, performance |
| Advanced TypeScript, type-safe frontend, tRPC | Frontend Engineer (TypeScript) | Type-safe React/Next.js projects |
| Visual design, typography, motion, aesthetics | Frontend Designer | Distinctive UI, design systems |
| Test strategy, E2E, coverage | QA Analyst | Testing architecture, automation |
| Monitoring, SLOs, performance, reliability | SRE | Observability, incident response |
---
## Choosing Between Generalist and Specialist Agents
### Backend Engineers
**Use `backend-engineer` (language-agnostic) when:**
- Language hasn't been decided yet
- Comparing multiple language options
- Multi-language system architecture
- You want recommendations on language choice
**Use language-specific engineers when:**
- Language is already decided
- You need framework-specific guidance (NestJS, FastAPI, etc.)
- Performance optimization for specific runtime
- Language-specific patterns and idioms
**Example decision:**
- "Should I use Go or TypeScript?" → **backend-engineer** (agnostic)
- "How do I optimize this Go service?" → **backend-engineer-golang**
- "Design a FastAPI endpoint" → **backend-engineer-python**
### Frontend Engineers
**Use `frontend-engineer` (general) when:**
- Standard React/Next.js projects
- TypeScript is used but not central concern
- Focus on component design, state, architecture
- General frontend best practices
**Use `frontend-engineer-typescript` when:**
- TypeScript is central to project success
- Need advanced type patterns (branded types, generics)
- Type-safe API integration (tRPC, Zod)
- Migrating JavaScript to TypeScript
- Strict type safety requirements
**Example decision:**
- "Build a React dashboard" → **frontend-engineer**
- "Build type-safe full-stack with tRPC" → **frontend-engineer-typescript**
---
## When to Use Developer Specialists vs General Review
### Use Developer Specialists for:
-**Deep technical expertise needed** Architecture decisions, complex implementations
-**Technology-specific guidance** "How do I optimize this Go service?"
-**Specialized domains** Infrastructure, SRE, testing strategy
-**Building from scratch** New service, new pipeline, new testing framework
### Use General Review Agents for:
-**Code quality assessment** Architecture, patterns, maintainability
-**Correctness & edge cases** Business logic verification
-**Security review** OWASP, auth, validation
-**Post-implementation** Before merging existing code
**Both can be used together:** Get developer specialist guidance during design, then run general reviewers before merge.
---
## Dispatching Multiple Specialists
If you need multiple specialists (e.g., backend engineer + DevOps engineer), dispatch in **parallel** (single message, multiple Task calls):
```
✅ CORRECT:
Task #1: ring-dev-team:backend-engineer-golang
Task #2: ring-dev-team:devops-engineer
(Both run in parallel)
❌ WRONG:
Task #1: ring-dev-team:backend-engineer-golang
(Wait for response)
Task #2: ring-dev-team:devops-engineer
(Sequential = 2x slower)
```
---
## ORCHESTRATOR Principle
Remember:
- **You're the orchestrator** Dispatch specialists, don't implement directly
- **Don't read specialist docs yourself** Dispatch to specialist, they know their domain
- **Combine with using-ring principle** Skills + Specialists = complete workflow
### Good Example (ORCHESTRATOR):
> "I need a Go service. Let me dispatch backend-engineer-golang to design it."
### Bad Example (OPERATOR):
> "I'll manually read Go best practices and design the service myself."
---
## Available in This Plugin
**Agents:**
- backend-engineer (language-agnostic)
- backend-engineer-golang
- backend-engineer-typescript
- backend-engineer-python
- devops-engineer
- frontend-engineer
- frontend-engineer-typescript
- frontend-designer
- qa-analyst
- sre
**Skills:**
- using-dev-team: Plugin introduction and agent selection guide
- writing-code: Developer agent selection and invocation patterns
**Note:** If a skill documents a developer agent but you can't find it, you may not have ring-dev-team enabled. Check `.claude-plugin/marketplace.json` or install ring-dev-team plugin.
---
## Integration with Other Plugins
- **using-ring** (default) ORCHESTRATOR principle for ALL agents
- **using-finops-team** Financial/regulatory agents
- **using-pm-team** Pre-dev workflow agents
Dispatch based on your need:
- General code review → default plugin agents
- Specific domain expertise → ring-dev-team agents
- Regulatory compliance → ring-finops-team agents
- Feature planning → ring-pm-team agents

View File

@@ -0,0 +1,398 @@
---
name: writing-code
description: |
Developer agent selection matrix - helps identify the best developer specialist
based on technology stack, domain, and requirements.
trigger: |
- Starting implementation work
- Need to choose between multiple developer agents
- Technology stack determines agent choice
skip_when: |
- Already know which agent to use → dispatch directly
- Not implementation work → use other skills
---
# Writing Code - Developer Agent Selection
## Purpose
This skill helps you identify and dispatch the most appropriate developer agent from the ring-dev-team plugin based on the task requirements.
## Available Developer Agents
| Agent | Specialization | Best For |
|-------|----------------|----------|
| `ring-dev-team:backend-engineer` | Language-agnostic backend (adapts to any language) | Unknown stack, polyglot systems, API design across languages |
| `ring-dev-team:backend-engineer-golang` | Go, APIs, microservices, databases, message queues | Go backend services, REST/gRPC APIs, database design, multi-tenancy |
| `ring-dev-team:backend-engineer-typescript` | TypeScript/Node.js, Express, Fastify, Prisma | Node.js backend, TypeScript APIs, full-stack TypeScript projects |
| `ring-dev-team:backend-engineer-python` | Python, FastAPI, Django, SQLAlchemy, Celery | Python backend, ML services, data pipelines, FastAPI/Django APIs |
| `ring-dev-team:frontend-engineer` | React, Next.js, TypeScript, state management | UI components, dashboards, forms, client-side logic |
| `ring-dev-team:frontend-engineer-typescript` | TypeScript-first React/Next.js, type-safe state | Type-heavy frontends, complex state, full-stack TypeScript |
| `ring-dev-team:frontend-designer` | Visual design, typography, UI aesthetics | Landing pages, marketing sites, design systems, distinctive interfaces |
| `ring-dev-team:devops-engineer` | CI/CD, Docker, Kubernetes, Terraform, IaC | Pipelines, containerization, infrastructure, deployment |
| `ring-dev-team:qa-analyst` | Test strategy, E2E, performance testing | Test plans, automation frameworks, quality gates |
| `ring-dev-team:sre` | Monitoring, observability, reliability | Alerting, SLOs, incident response, performance |
## Decision Matrix
### By Technology Stack
```
Backend work?
├── Go/Golang code?
│ └── Yes → backend-engineer-golang
├── TypeScript/Node.js backend?
│ └── Yes → backend-engineer-typescript
├── Python backend (FastAPI/Django)?
│ └── Yes → backend-engineer-python
├── Unknown/multiple languages?
│ └── Yes → backend-engineer (language-agnostic)
└── Frontend work?
├── React/Next.js?
│ ├── TypeScript-heavy with complex state?
│ │ └── Yes → frontend-engineer-typescript
│ ├── Visual design/UI aesthetics focus?
│ │ └── Yes → frontend-designer
│ └── Component logic/functionality focus?
│ └── Yes → frontend-engineer
├── Infrastructure/CI-CD/Docker?
│ └── Yes → devops-engineer
├── Testing/QA/Automation?
│ └── Yes → qa-analyst
└── Monitoring/Reliability/Performance?
└── Yes → sre
```
### By Task Type
| Task | Primary Agent | Supporting Agent |
|------|---------------|------------------|
| Build REST API (Go) | backend-engineer-golang | qa-analyst (tests) |
| Build REST API (TypeScript) | backend-engineer-typescript | qa-analyst (tests) |
| Build REST API (Python) | backend-engineer-python | qa-analyst (tests) |
| Build REST API (unknown language) | backend-engineer | qa-analyst (tests) |
| Create UI component | frontend-engineer | qa-analyst (E2E) |
| Create type-safe UI component | frontend-engineer-typescript | qa-analyst (E2E) |
| Design landing page | frontend-designer | frontend-engineer (implementation) |
| Full-stack TypeScript feature | backend-engineer-typescript | frontend-engineer-typescript |
| Set up CI/CD | devops-engineer | sre (monitoring) |
| Write test suite | qa-analyst | - |
| Add monitoring | sre | devops-engineer (infra) |
| Database schema (Go) | backend-engineer-golang | - |
| Database schema (TypeScript) | backend-engineer-typescript | - |
| Database schema (Python) | backend-engineer-python | - |
| Deploy to K8s | devops-engineer | sre (observability) |
| Performance optimization | sre | backend-engineer-* |
## Agent Dispatch Pattern
### Single Agent Task
```
Task tool:
subagent_type: "ring-dev-team:backend-engineer-golang"
prompt: "Design and implement a REST API for user authentication with JWT tokens"
```
### Multi-Agent Workflow
For complex features requiring multiple specialists:
```
Step 1: Backend API
Task tool → ring-dev-team:backend-engineer-golang
"Implement the user service with CRUD operations"
Step 2: Frontend Integration
Task tool → ring-dev-team:frontend-engineer
"Create React components to consume the user API"
Step 3: Testing
Task tool → ring-dev-team:qa-analyst
"Write E2E tests for the user registration flow"
Step 4: Deployment
Task tool → ring-dev-team:devops-engineer
"Create Kubernetes manifests and CI/CD pipeline"
Step 5: Observability
Task tool → ring-dev-team:sre
"Add monitoring, alerting, and SLO definitions"
```
## Agent Capabilities Detail
### backend-engineer
**Technologies:**
- Language-agnostic (adapts to Go, TypeScript, Python, Java, Rust, etc.)
- RESTful API design principles
- Database design (SQL and NoSQL)
- Message queues and event streaming
- Authentication and authorization patterns
**Patterns:**
- Clean Architecture / Hexagonal
- Repository pattern
- Dependency injection
- SOLID principles
- API design best practices
**Best practices:**
- Language-appropriate error handling
- Security best practices (OWASP)
- Scalability and performance
- Testing strategies
- Documentation standards
**When to use:**
- Codebase uses unknown or multiple languages
- Need language-agnostic architecture advice
- Polyglot systems requiring consistency
- Initial API design before language selection
### backend-engineer-golang
**Technologies:**
- Go (Fiber, Gin, Echo frameworks)
- PostgreSQL, MongoDB, Redis
- RabbitMQ, Kafka
- gRPC, REST
- OAuth2, JWT, WorkOS
**Patterns:**
- Clean Architecture / Hexagonal
- Repository pattern
- Dependency injection
- Multi-tenancy (schema-per-tenant, row-level)
**Best practices:**
- Error handling (no panic in production)
- Context propagation
- Graceful shutdown
- Connection pooling
### backend-engineer-typescript
**Technologies:**
- TypeScript, Node.js
- Express, Fastify, NestJS
- Prisma, TypeORM, Sequelize
- PostgreSQL, MongoDB, Redis
- Bull, BullMQ (job queues)
- JWT, Passport.js, OAuth2
**Patterns:**
- Layered architecture
- Dependency injection (NestJS, tsyringe)
- Repository pattern
- DTO validation (class-validator, Zod)
- Middleware patterns
**Best practices:**
- Strict TypeScript configuration
- Async/await error handling
- Type-safe database queries
- Request validation
- Graceful shutdown
- Logging and observability
### backend-engineer-python
**Technologies:**
- Python 3.10+
- FastAPI, Django, Flask
- SQLAlchemy, Django ORM
- PostgreSQL, MongoDB, Redis
- Celery, RQ (task queues)
- Pydantic (validation)
- pytest, unittest
**Patterns:**
- MVC / MVT (Django)
- Dependency injection (FastAPI)
- Repository pattern
- Service layer pattern
- Type hints and validation
**Best practices:**
- Type hints everywhere
- Exception handling
- Virtual environments
- Async/await (asyncio)
- Database migrations (Alembic)
- API documentation (OpenAPI)
### frontend-engineer
**Technologies:**
- React 18+, Next.js 14+
- TypeScript
- Tailwind CSS, Styled Components
- React Query, Zustand, Redux
- Jest, React Testing Library, Playwright
**Patterns:**
- Component composition
- Custom hooks
- Server components (Next.js)
- Optimistic updates
**Best practices:**
- Accessibility (WCAG 2.1)
- Performance (Core Web Vitals)
- Type safety
- Error boundaries
### frontend-engineer-typescript
**Technologies:**
- TypeScript (strict mode)
- React 18+, Next.js 14+
- Zod, io-ts (runtime validation)
- tRPC (end-to-end type safety)
- TanStack Query (React Query)
- Zustand, Jotai (type-safe state)
- Vitest, Jest with type coverage
**Patterns:**
- Type-driven development
- Generic components with type constraints
- Discriminated unions for state
- Type-safe API clients
- Schema-driven forms
**Best practices:**
- No `any` types (strict mode)
- Type guards and narrowing
- Exhaustive switch statements
- Runtime validation at boundaries
- Type-safe routing
- Inference over explicit types
### frontend-designer
**Technologies:**
- Figma, Sketch design systems
- Tailwind CSS, CSS-in-JS
- Typography systems (Google Fonts, custom fonts)
- Animation libraries (Framer Motion, GSAP)
- Color theory and accessibility tools
**Patterns:**
- Visual hierarchy
- Responsive design principles
- Design tokens
- Component variants
**Best practices:**
- Accessible color contrast (WCAG AA/AAA)
- Responsive typography scales
- Consistent spacing systems
- Performance-optimized assets
- Brand consistency
### devops-engineer
**Technologies:**
- Docker, Docker Compose
- Kubernetes, Helm
- Terraform, Pulumi
- GitHub Actions, GitLab CI
- AWS, GCP, Azure
**Patterns:**
- GitOps
- Infrastructure as Code
- Blue-green / Canary deployments
- Secret management
**Best practices:**
- Immutable infrastructure
- Least privilege
- Automated rollbacks
- Cost optimization
### qa-analyst
**Technologies:**
- Jest, Vitest, Go testing
- Playwright, Cypress
- k6, Artillery (load testing)
- Postman, Newman
**Patterns:**
- Test pyramid
- Page Object Model
- Data-driven testing
- Contract testing
**Best practices:**
- Shift-left testing
- Test isolation
- Deterministic tests
- CI integration
### sre
**Technologies:**
- Prometheus, Grafana
- Datadog, New Relic
- PagerDuty, OpsGenie
- Jaeger, OpenTelemetry
**Patterns:**
- SLI/SLO/SLA definitions
- Error budgets
- Incident management
- Chaos engineering
**Best practices:**
- Alerting hygiene
- Runbooks
- Post-mortems
- Capacity planning
## When NOT to Use Developer Agents
- **Code review** → Use `ring-default:code-reviewer` (parallel with business-logic and security)
- **Planning/Design** → Use `ring-default:brainstorming` or `ring-default:write-plan`
- **Debugging** → Use `ring-default:systematic-debugging` skill
- **Architecture exploration** → Use `ring-default:codebase-explorer`
## Integration with Ring Workflows
### With TDD (test-driven-development)
```
1. Use qa-analyst to design test cases
2. Use appropriate developer agent to implement
3. Use qa-analyst to verify test coverage
```
### With Code Review (requesting-code-review)
```
1. Use developer agent to implement
2. Use ring-default reviewers (parallel) to review
3. Use developer agent to address findings
```
### With Pre-Dev Planning
```
1. Use pm-team skills for planning (PRD, TRD)
2. Use ring-default:write-plan for implementation plan
3. Use this skill to identify developer agents for each task
4. Dispatch agents per task with code review between
```
## Remember
1. **Match agent to task** - Don't use backend agent for frontend work
2. **Combine agents for full-stack** - Complex features need multiple specialists
3. **Always review** - Use code reviewers after implementation
4. **Follow skill workflows** - TDD, systematic debugging, verification
5. **Document decisions** - Agents can explain their choices