Ruflow -- Complete Training Guide for Claude Code

Ruflow v3.5 (formerly "Claude Flow") -- 5,900+ commits, 55 alpha iterations, 259 MCP tools, 60+ agents, 8 AgentDB controllers. Packages: @claude-flow/cli@3.5.0, claude-flow@3.5.0, ruflo@3.5.0

This is NOT a typical project README. This is a training manual. Any Claude Code instance that opens this directory gets immediate access to the entire Ruflow skill, agent, and asset library described below.

1. What Is Ruflow
2. Directory Structure Map
3. Complete Skill Catalog (71 Skills)
4. Agent Skills Library (1,521 Skills)
5. Commands Reference (172 Commands)
6. Source Repositories in Detail
7. V3 Packages (@claude-flow)
8. Assets and Resources Available
9. Configuration Reference
10. When to Use What (Decision Matrix)
11. Swarm Orchestration Guide
12. Hooks System

Ruflow is a training directory that equips any Claude Code instance with a massive library of capabilities the moment the directory is opened. It is not a single application -- it is an orchestration layer, skill corpus, and toolchain combined.

When Claude Code opens this directory, it automatically loads:

• CLAUDE.md (39KB of project instructions) via the project config system
• 71 invocable skills from .claude/skills/
• 172 slash commands from .claude/commands/
• 34 agent definitions from .claude/agents/
• 40 background helpers from .claude/helpers/
• MCP server connections from .claude/mcp.json and .claude/settings.json

The 1,521 agent skills in .agents/skills/ are available for reference and can be loaded on demand. The 7 source repositories in src/ provide raw skill data, CLI tools, memory systems, and content pipelines.

Ruflow/
|
|-- .claude/                           # Claude Code configuration system
|   |-- skills/ (71 skills)            # Directly invocable skills (see Section 3)
|   |-- commands/ (172 commands)       # Slash commands organized by domain (see Section 5)
|   |-- agents/ (34 agents)            # Agent type definitions with SKILL.md files
|   |-- helpers/ (40 helpers)          # Background hooks, workers, and automation scripts
|   |-- config/                        # Performance and dependency configs
|   |   |-- v3-dependency-optimization.json
|   |   +-- v3-performance-targets.json
|   |-- checkpoints/                   # Session checkpoint storage
|   |-- settings.json                  # Main project settings (model, permissions, hooks)
|   |-- settings.local.json            # Local overrides
|   |-- mcp.json                       # MCP server configuration
|   |-- statusline.mjs                 # Status line display script
|   |-- statusline.sh                  # Shell status line
|   +-- statusline-command.sh          # Status line command handler
|
|-- .agents/
|   +-- skills/ (1,521 skills)         # Full agent skill library (see Section 4)
|
|-- src/                               # Source repositories (7 repos)
|   |-- antigravity-awesome-skills/    # 8,418 files -- skill corpus source
|   |-- claude-mem/                    # 24,921 files -- persistent memory system
|   |-- awesome-claude-code/           # 681 files -- curated tools and resources
|   |-- ui-ux-pro-max-skill/          # 964 files -- UI/UX design CLI and data
|   |-- superpowers/                   # 124 files -- workflow skills and commands
|   |-- koda-stack/                    # 13 files -- video production pipeline
|   +-- website-builder-setup/         # 2 files -- web builder installer
|
|-- v3/                                # V3 monorepo
|   |-- @claude-flow/ (21 packages)    # Core packages (see Section 7)
|   |-- agents/                        # V3 agent system
|   |-- helpers/                       # V3 helper scripts
|   |-- implementation/                # Implementation tracking
|   |-- mcp/                           # MCP server source
|   |-- plugins/                       # Plugin system
|   |-- scripts/                       # Build and utility scripts
|   |-- src/                           # V3 core source
|   |-- __tests__/                     # V3 test suite
|   +-- docs/                          # V3 documentation
|
|-- v2/                                # V2 stable production code
|-- ruflo/                             # Ruflo SDK/CLI wrapper package
|-- bin/                               # CLI entry points (cli.js)
|-- scripts/                           # Build, utility, and memory-db scripts
|-- docs/                              # Documentation
|-- tests/                             # Test suite
|-- agents/                            # Legacy agent definitions
|-- plugin/                            # Plugin source
|
|-- CLAUDE.md                          # Project instructions (39KB) -- loaded automatically
|-- CLAUDE.local.md                    # Local development config
|-- AGENTS.md                          # Agent reference documentation
|-- package.json                       # Main package config (v3.5.48)
|-- tsconfig.json                      # TypeScript configuration
|-- CHANGELOG.md                       # Release history
|-- SECURITY.md                        # Security policy
+-- LICENSE                            # License file

All 71 skills live in .claude/skills/ and each contains a SKILL.md file that Claude Code loads when the skill is invoked. Skills are grouped below by function.

These 10 skills form a complete video content creation pipeline. Execute them in order:

Source: src/koda-stack/skills/

Source: src/superpowers/skills/

The .agents/skills/ directory contains 1,521 pre-built agent skills. Each skill is a directory with a SKILL.md file defining the agent's capabilities, instructions, and domain knowledge. These cover virtually every technology domain.

Agent skills are loaded by reading their SKILL.md file. To use a skill:

# Read the skill definition
Read .agents/skills/<skill-name>/SKILL.md

# Follow the instructions in the SKILL.md to apply the skill's capabilities

The skill definitions contain detailed instructions, patterns, code templates, and domain knowledge that Claude Code can follow to perform specialized tasks.

Commands live in .claude/commands/ and are invoked as slash commands. There are 8 top-level command files and 19 command directories containing 164 sub-commands.

agent-capabilities, agent-coordination, agent-spawning, agent-types, README

bottleneck-detect, COMMAND_COMPLIANCE_REPORT, performance-bottlenecks, performance-report, README, token-efficiency, token-usage

auto-agent, README, self-healing, session-memory, smart-agents, smart-spawn, workflow-select

agent-spawn, init, orchestrate, README, spawn, swarm-init, task-orchestrate

app-store, challenges, login-registration, neural-network, payments, sandbox, swarm, user-tools, workflow

code-review, code-review-swarm, github-modes, github-swarm, issue-tracker, issue-triage, multi-repo-swarm, pr-enhance, pr-manager, project-board-sync, README, release-manager, release-swarm, repo-analyze, repo-architect, swarm-issue, swarm-pr, sync-coordinator, workflow-automation

hive-mind, hive-mind-consensus, hive-mind-init, hive-mind-memory, hive-mind-metrics, hive-mind-resume, hive-mind-sessions, hive-mind-spawn, hive-mind-status, hive-mind-stop, hive-mind-wizard, README

overview, post-edit, post-task, pre-edit, pre-task, README, session-end, setup

memory-persist, memory-search, memory-usage, neural, README

agent-metrics, agents, README, real-time-view, status, swarm-monitor

auto-topology, cache-manage, parallel-execute, parallel-execution, README, topology-optimize

commands, config, examples, modes, README, session, start

analyzer, architect, ask, batch-executor, code, coder, debug, debugger, designer, devops, docs-writer, documenter, innovator, integration, mcp, memory-manager, optimizer, orchestrator, post-deployment-monitoring-mode, refinement-optimization-mode, researcher, reviewer, security-review, sparc, sparc-modes, spec-pseudocode, supabase-admin, swarm-coordinator, tdd, tester, tutorial, workflow-manager

pipeline, run

analysis, development, examples, maintenance, optimization, README, research, swarm, swarm-analysis, swarm-background, swarm-init, swarm-modes, swarm-monitor, swarm-spawn, swarm-status, swarm-strategies, testing

model-update, neural-patterns, neural-train, pattern-learn, README, specialization

start

check, start

development, README, research, workflow-create, workflow-execute, workflow-export

Seven source repositories live in src/, totaling 35,123 files. Each serves a distinct purpose.

The most comprehensive UI/UX design toolkit. Contains a full CLI application built with TypeScript, commander, chalk, and ora.

Structure:

• cli/ -- CLI application source
• src/ui-ux-pro-max/ -- Core skill data and scripts
• docs/ -- Documentation
• preview/ and screenshots/ -- Visual previews

Data Assets (60 CSVs):

14 general design data CSVs (each appears in both cli/assets/data/ and src/ui-ux-pro-max/data/):

16 stack-specific CSVs (in cli/assets/data/stacks/):

angular, astro, flutter, html-tailwind, jetpack-compose, laravel, nextjs, nuxtjs, nuxt-ui, react, react-native, shadcn, svelte, swiftui, threejs, vue

Python Scripts (4):

• _sync_all.py -- Sync data across directories
• core.py -- Core design system operations
• design_system.py -- Design system generation
• search.py -- Design data search

When to use: Any UI/UX design work, component generation, design system creation, landing page design, or stack-specific UI patterns.

A structured development workflow toolkit providing 14 skills, 3 commands, and 1 agent.

Structure:

• skills/ -- 14 workflow skills (see Development Workflow Skills above)
• commands/ -- brainstorm.md, execute-plan.md, write-plan.md
• agents/ -- code-reviewer.md
• hooks/ -- Git and session hooks
• scripts/ -- Utility scripts
• tests/ -- Test suite
• CLAUDE-SUPERPOWERS.md -- Full superpowers documentation
• GEMINI.md -- Gemini adapter configuration

When to use: Planning features, conducting code reviews, TDD, systematic debugging, brainstorming, Git workflow management, or parallel agent dispatch.

A persistent memory system that survives across Claude Code sessions. The largest source repo by file count due to its dependency tree.

Structure:

• src/ -- Core memory system source
• scripts/ -- Memory management scripts
• conductor.json -- Conductor configuration
• cursor-hooks/ -- Cursor IDE adapter
• docs/ -- Memory system documentation
• tests/ -- Test suite
• CLAUDE-CLAUDEMEM.md -- Full memory system documentation

Key Features:

• Multi-tier storage architecture (short-term, long-term, episodic)
• CLI adapters for claude-code, cursor, gemini, windsurf
• MCP server integration for cross-session memory
• UI viewer with fonts and assets

When to use: Cross-session memory persistence, storing decisions and context that should survive conversation resets, building agents that learn over time.

A curated resource library for Claude Code tooling, templates, and best practices.

Structure:

• THE_RESOURCES_TABLE.csv -- 222 curated entries of Claude Code tools and resources
• scripts/ -- 15 utility scripts
• data/ -- Resource data
• templates/ -- Project and configuration templates
• tools/ -- Utility tools
• resources/ -- Reference materials
• assets/ -- Static assets
• docs/ -- Documentation

When to use: Finding Claude Code resources, looking up community tools, referencing best practices, or using project templates.

The source corpus for the 1,521 agent skills in .agents/skills/. Contains the raw skill definitions, indexes, and metadata.

Structure:

• skills/ -- Raw skill source files
• data/ -- Skill metadata and indexes
• docs/ -- Skill documentation
• scripts/ -- Build and sync scripts
• tools/ -- Skill management tools
• plugins/ -- Plugin integrations

All "antigravity" references in the source have been renamed to "claude-code" for consistency.

When to use: Referencing the raw source of agent skills, understanding skill structure, or building new skills from templates.

A minimal installer skill that sets up a 4-tool web development stack:

1. UI/UX Pro Max -- Full design system
2. Framer Motion -- Animation library
3. 21st.dev Magic -- AI component generation
4. shadcn/ui -- Component library

When to use: Quick scaffolding of a new website project with a complete design and component stack.

A creative video production pipeline with 10 specialized agents.

Structure:

• skills/ -- 10 production skills (brief, trends, concept, script, art-direction, storyboard, generate, assemble, publish, repurpose)
• CLAUDE-KODA.md -- Creative DNA configuration for tone, style, and brand voice
• README.md -- Usage guide

When to use: Video content creation, from initial brief through final publication and repurposing.

The v3/@claude-flow/ directory contains 21 packages that make up the V3 monorepo:

Location: .claude/skills/ui-styling/canvas-fonts/

54 TTF font files available for canvas rendering, image generation, and design work. These are loaded by the ui-styling skill.

Location: src/ui-ux-pro-max-skill/cli/assets/data/

• 14 general design reference CSVs (colors, typography, icons, charts, etc.)
• 16 stack-specific CSVs (angular, react, nextjs, flutter, svelte, vue, etc.)
• Duplicated in src/ui-ux-pro-max-skill/src/ui-ux-pro-max/data/ for the core skill

Location: .claude/skills/brand/scripts/

The website-builder-setup skill installs:

• Framer Motion -- React animation library
• 21st.dev Magic -- AI component generation
• shadcn/ui -- Accessible component library
• UI/UX Pro Max -- Complete design system

Location: src/awesome-claude-code/THE_RESOURCES_TABLE.csv

222 curated Claude Code tools, templates, and resources indexed by category.

Location: scripts/memory-db/ (build scripts), runtime at data/memory/

SQLite + HNSW vector index for semantic memory search. Uses WASM sql.js for cross-platform compatibility.

Available via MCP server connections defined in .claude/settings.json:

Main project configuration. Key sections:

{
  "model": "claude-opus-4-6",
  "customInstructions": "Follow CLAUDE.md guidelines...",
  "env": {
    "CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS": "1",
    "CLAUDE_FLOW_V3_ENABLED": "true",
    "CLAUDE_FLOW_HOOKS_ENABLED": "true"
  },
  "permissions": {
    "allow": ["Bash(npx claude-flow*)", "mcp__claude-flow__*", "..."],
    "deny": ["Read(./.env)", "Read(./.env.*)", "Bash(rm -rf /)"]
  },
  "hooks": {
    "PreToolUse": ["..."],
    "PostToolUse": ["..."]
  }
}

MCP server connections. Currently configured:

• flow-nexus -- Supabase-backed MCP server (requires SUPABASE_URL and SUPABASE_ANON_KEY)

The main MCP servers (claude-flow, ruv-swarm) are started via npx commands defined in permissions.

The primary project instructions file. Loaded automatically by Claude Code. Contains:

• Behavioral rules
• File organization requirements
• Project architecture
• Concurrency patterns (batch all operations)
• Swarm orchestration protocols
• Agent routing tables
• V3 CLI command reference (26 commands, 140+ subcommands)
• Publishing procedures
• Plugin registry maintenance

Local development configuration with environment variables, quick references, and local overrides. Not committed to git.

Name: claude-flow
Version: 3.5.48
Type: ES module
Main: dist/index.js
Bin: ./bin/cli.js

Check .agents/skills/ for a dedicated skill. Examples:

Ruflow uses a multi-agent swarm system where MCP tools handle coordination and Claude Code's Task tool handles execution.

Topology:    hierarchical (central coordinator prevents drift)
Max Agents:  8 (smaller team = less drift)
Strategy:    specialized (clear roles, no overlap)
Consensus:   raft (leader maintains authoritative state)
Memory:      hybrid (SQLite + AgentDB)

// Initialize swarm via MCP
mcp__claude-flow__swarm_init({
  topology: "hierarchical",
  maxAgents: 8,
  strategy: "specialized"
})

Codes 1-11 use hierarchical/specialized topology. Code 13 uses mesh/balanced.

Use swarm for:

• Multi-file changes (3+ files)
• New feature implementation
• Refactoring across modules
• API changes with tests
• Security-related changes
• Performance optimization
• Database schema changes

Skip swarm for:

• Single file edits
• Simple bug fixes (1-2 lines)
• Documentation updates
• Configuration changes
• Quick questions

Location: .claude/helpers/

# Initialize project
npx claude-flow@v3alpha init --wizard

# Start daemon
npx claude-flow@v3alpha daemon start

# Spawn an agent
npx claude-flow@v3alpha agent spawn -t coder --name my-coder

# Initialize swarm
npx claude-flow@v3alpha swarm init --v3-mode

# Search memory
npx claude-flow@v3alpha memory search -q "pattern query"

# Run diagnostics
npx claude-flow@v3alpha doctor --fix

# Security scan
npx claude-flow@v3alpha security scan --depth full

# Performance benchmark
npx claude-flow@v3alpha performance benchmark --suite all

# Hook operations
npx claude-flow@v3alpha hooks pre-task --description "task"
npx claude-flow@v3alpha hooks post-task --task-id "id" --success true
npx claude-flow@v3alpha hooks session-start --session-id "id"

# Worker management
npx claude-flow@v3alpha hooks worker list
npx claude-flow@v3alpha hooks worker dispatch --trigger audit

• Repository: https://github.com/ruvnet/claude-flow
• Issues: https://github.com/ruvnet/claude-flow/issues
• Funding: https://github.com/sponsors/ruvnet