Agents

Agents

Ruflo ships with 60+ specialized agent types. Each role has a distinct mission in the team.

---

Core Development

Agent	Purpose
coder	Write and refactor code
reviewer	Code quality, style, and design review
tester	Test generation and validation
planner	Task decomposition and planning
researcher	Investigation and analysis

V3 Specialized Agents

Agent	Purpose
security-architect	Security design and threat modeling
security-auditor	Vulnerability scanning and compliance
memory-specialist	Agent memory and knowledge management
performance-engineer	Optimization and profiling

Swarm Coordination

Agent	Purpose
hierarchical-coordinator	Central queue and task assignment
mesh-coordinator	Peer-to-peer message routing
adaptive-coordinator	Dynamic topology switching
collective-intelligence-coordinator	Consensus-driven decisions
swarm-memory-manager	Shared state and journaling

Consensus & Distributed

Agent	Purpose
byzantine-coordinator	BFT consensus (tolerates f < n/3 faults)
raft-manager	Leader-based consensus (tolerates f < n/2)
gossip-coordinator	Eventual consistency via epidemic protocol
consensus-builder	Multi-algorithm orchestration
crdt-synchronizer	Conflict-free replicated data types
quorum-manager	Configurable quorum voting
security-manager	Byzantine fault detection

Performance & Optimization

Agent	Purpose
perf-analyzer	Profile and identify bottlenecks
performance-benchmarker	Run and compare benchmarks
task-orchestrator	Load distribution and scheduling
memory-coordinator	Cache optimization and GC tuning
smart-agent	Cost-aware routing

GitHub & Repository

Agent	Purpose
github-modes	GitHub API integration
pr-manager	Pull request lifecycle
code-review-swarm	Coordinated PR reviews
issue-tracker	Issue triage and assignment
release-manager	Version and release automation
workflow-automation	CI/CD pipeline control
project-board-sync	GitHub Projects sync
repo-architect	Repository structure analysis
multi-repo-swarm	Cross-repo coordination

SPARC Methodology

Agent	Purpose
sparc-coord	Orchestrate SPARC phases
sparc-coder	Code generation in SPARC
specification	Requirements and specs
pseudocode	Algorithm design
architecture	System design
refinement	Optimization and hardening

Specialized Development

Agent	Purpose
backend-dev	REST APIs, databases, services
mobile-dev	iOS/Android native development
ml-developer	ML pipeline and model training
cicd-engineer	CI/CD pipeline design
api-docs	API documentation generation
system-architect	High-level system design
code-analyzer	Static analysis and metrics
base-template-generator	Scaffold new projects

Testing & Validation

Agent	Purpose
tdd-london-swarm	Test-driven development (London School)
production-validator	Pre-release validation

---

How to Spawn an Agent

npx ruflo@latest agent spawn -t coder --name my-dev

Via Claude Code:

Task({
  prompt: "Implement the user authentication module",
  subagent_type: "coder",
  name: "auth-dev",
  run_in_background: true
})

---

Agent Routing (Anti-Drift)

Code	Task	Agents
1	Bug Fix	coordinator, researcher, coder, tester
3	Feature	coordinator, architect, coder, tester, reviewer
5	Refactor	coordinator, architect, coder, reviewer
7	Performance	coordinator, perf-engineer, coder
9	Security	coordinator, security-architect, auditor
11	Memory/DB	coordinator, memory-specialist, perf-engineer
13	Docs	researcher, api-docs

---

Agent Lifecycle

# Spawn
npx ruflo@latest agent spawn -t coder --name my-agent

# List
npx ruflo@latest agent list

# Status
npx ruflo@latest agent status <agent-id>

# Logs
npx ruflo@latest agent logs <agent-id> --last 10

# Metrics
npx ruflo@latest agent metrics <agent-id>

# Stop
npx ruflo@latest agent stop <agent-id>

---

Teams & Communication

Agents use SendMessage to communicate within a team:

// Architect designs, then tells Developer
SendMessage({
  to: "developer",
  summary: "REST API design",
  message: "[design document]"
})

// Developer acknowledges and reports progress
SendMessage({
  to: "architect",
  summary: "Implementation started",
  message: "Endpoints in progress..."
})

---

Customization

Create custom agents via plugins or skill definitions in CLAUDE.md.

---

Ruflo v3.10.1 · GitHub