UNPKG

claude-flow-novice

Version:

Claude Flow Novice - Advanced orchestration platform for multi-agent AI workflows with CFN Loop architecture Includes Local RuVector Accelerator and all CFN skills for complete functionality.

219 lines (162 loc) 6.45 kB
# CFN Dev Team Agent Structure ## Overview The CFN (Claude Flow Novice) development team comprises 23 production agents organized into 4 strategic categories, designed to provide comprehensive software development and workflow management capabilities. ## Directory Structure ### Coordinators - **Purpose**: CFN Loop orchestration and workflow management - **Key Agents**: - `cfn-v3-coordinator`: Primary workflow orchestrator ### Developers - **Purpose**: Core implementation and creative problem-solving - **Key Agents**: - `coder`: Production code implementation - `backend-dev`: Backend system design and implementation - `researcher`: Technical research and solution exploration - `architect`: System design and architectural planning - `agent-builder`: Agent template and workflow design ### Reviewers - **Purpose**: Quality assurance and code validation - **Key Agents**: - `reviewer`: Code review and quality assessment - `code-analyzer`: Static code analysis - `code-quality-validator`: Comprehensive code quality checks - `security-specialist`: Security vulnerability detection ### Testers - **Purpose**: Comprehensive testing and validation - **Key Agents**: - `tester`: General test strategy and implementation - `playwright-tester`: Web interaction testing - `interaction-tester`: User interaction validation - `production-validator`: Production readiness checks - `perf-analyzer`: Performance testing and optimization ## Agent Selection Guide ### When to Use Each Category 1. **Coordinators** - Complex workflow orchestration - Multi-agent collaboration scenarios - CFN Loop management 2. **Developers** - Initial implementation - Feature development - Technical problem-solving - Prototype creation 3. **Reviewers** - Post-implementation code review - Quality gate validation - Security and performance analysis - Refactoring recommendations 4. **Testers** - Comprehensive test strategy - Automated testing - Production validation - Performance optimization ## CFN Loop Integration ### Loop Participation Levels - **Loop 2**: Preliminary design and research - **Loop 3**: Implementation and initial validation - **Loop 4**: Advanced testing and production readiness ## Naming Conventions - All agents follow the `cfn-dev-team` namespace - Naming format: `category-specific-role` - Example: `developer-backend-specialist` ## Agent Template Requirements Each agent MUST include: 1. **Name**: Unique, descriptive identifier 2. **Description**: Clear purpose and capabilities 3. **Tools**: Permitted interaction tools 4. **Model**: Assigned AI model 5. **Capabilities**: Specific functional areas 6. **Lifecycle Hooks**: - SQLite tracking - Redis coordination 7. **ACL Level**: Access control level (1-5) ## Multi-Worktree Coordination ### Environment Variables Provided by Coordinator When coordinators spawn agents in multi-worktree environments, the following environment variables are automatically injected: ```bash COMPOSE_PROJECT_NAME="cfn-feature-auth" # Unique project name per branch CFN_REDIS_PORT=6421 # Redis port (6379 + offset) CFN_POSTGRES_PORT=5474 # Postgres port (5432 + offset) WORKTREE_BRANCH="feature-auth" # Git branch name ``` ### Service Discovery Pattern Agents use Docker service names for internal communication: ```bash # Redis connection (within Docker network) redis-cli -h redis -p 6379 # PostgreSQL connection (within Docker network) PGHOST=postgres PGPORT=5432 psql -U postgres # HTTP connections curl http://orchestrator:3001/health ``` **Important**: Service names are resolved by Docker's internal DNS within the network. Container names (e.g., `cfn-redis-1`) will NOT resolve. ### Multi-Worktree Examples #### Scenario 1: Feature Development (Feature-Auth Branch) ```bash # Branch: feature-auth # Offset: ~42 (calculated from branch name) # Ports: Redis=6421, Postgres=5474, Orchestrator=3043 # Coordinator injects: export COMPOSE_PROJECT_NAME="cfn-feature-auth" export CFN_REDIS_PORT=6421 export CFN_POSTGRES_PORT=5474 # Agent connects to correct worktree services redis-cli -h redis -p 6421 psql -h postgres -p 5474 ``` #### Scenario 2: Bugfix Work (Bugfix-Validation Branch) ```bash # Branch: bugfix-validation # Offset: ~78 (calculated from branch name) # Ports: Redis=6457, Postgres=5510, Orchestrator=3079 # Coordinator injects: export COMPOSE_PROJECT_NAME="cfn-bugfix-validation" export CFN_REDIS_PORT=6457 export CFN_POSTGRES_PORT=5510 # Agent connects to correct worktree services redis-cli -h redis -p 6457 psql -h postgres -p 5510 ``` #### Scenario 3: Main Branch (Production Ready) ```bash # Branch: main/master # Offset: 0 (main gets priority) # Ports: Redis=6379, Postgres=5432, Orchestrator=3001 # Coordinator injects: export COMPOSE_PROJECT_NAME="cfn-main" export CFN_REDIS_PORT=6379 export CFN_POSTGRES_PORT=5432 # Standard ports - no offset redis-cli -h redis -p 6379 psql -h postgres -p 5432 ``` ### Running Multiple Worktrees Simultaneously Team members can develop in parallel without conflicts: ```bash # Developer 1: Feature branch worktree_1/feature-auth$ ./scripts/docker/run-in-worktree.sh up -d # Ports: Redis=6421, Postgres=5474 # Developer 2: Bugfix branch (same machine) worktree_2/bugfix-validation$ ./scripts/docker/run-in-worktree.sh up -d # Ports: Redis=6457, Postgres=5510 # Developer 3: Main branch worktree_3/main$ ./scripts/docker/run-in-worktree.sh up -d # Ports: Redis=6379, Postgres=5432 # All three run simultaneously without port conflicts! ``` ## Adding New Agents ### Process 1. Use `agent-builder` for initial template creation 2. Follow agent template structure 3. Validate against CFN Loop coordination patterns 4. Submit for team review 5. Integrate into appropriate category ### Validation Checklist - [ ] Unique name and purpose - [ ] Defined capabilities - [ ] Appropriate tool selection - [ ] Lifecycle hook configuration - [ ] ACL level assignment - [ ] Category alignment ## Contributing Agents are critical to our workflow. Propose new agents or improvements via pull request to the CFN development team.