Files
Dragan Spiridonov c77d1b5062 fix(hooks): apply the Windows argv fix from #3322 to the other three shim copies (#3332)
#3322 escaped argv on the Windows shell:true path in
plugins/ruflo-core/scripts/ruflo-hook.cjs. The same shim exists in three
other places, which still pass hook-derived values (a Bash tool's
`command`, a file path) to cmd.exe unescaped, and are broader than the
copy that was fixed — a bare `process.platform === 'win32'` with no
exemption for `node`:

  - .claude-plugin/scripts/ruflo-hook.cjs   (published in the npm package)
  - plugin/scripts/ruflo-hook.cjs           (byte-identical sibling)
  - generateRufloHookCjs() in helpers-generator.ts, which `ruflo init`
    writes to .claude/helpers/ruflo-hook.cjs

All four now follow one pattern — resolve, then escape:

  - resolveCommandPath() walks PATH/PATHEXT with fs only. The previous
    probe was `execSync('where ' + cmd)`, which spawned a shell on every
    hook invocation.
  - resolveNpmShim() maps an npm .cmd shim to the .js entrypoint it would
    have run, read from the package's own `bin` field rather than a
    guessed filename, and required to resolve inside the package
    directory. Both npm layouts are handled (global prefix and
    node_modules/.bin), as is npx, whose command and package names differ.
  - invokeHook() spawns `node <entry>` with shell:false, so CreateProcess
    receives the argv array verbatim — no second cmd.exe tokenizer, and no
    %VAR% expansion, which quoting does not suppress and carets do not
    reliably escape.
  - escapeCmdArg() is unchanged from #3322 and remains as the fallback for
    the case where no entrypoint can be identified. Returning early there
    would silently drop the hook instead of surfacing the problem.

escapeCmdArg, resolveCommandPath, resolveNpmShim and resolveInvocation are
byte-identical across all four copies, and the suite asserts that for each
of them — the divergence between copies is what let this persist after
#3322.

Verification. #3322 could only assert the string transform, with no
Windows host available. Two changes address that:

  - resolveCommandPath/resolveInvocation take `platform` and `env` as
    arguments, so the Windows branch runs on any OS. The tests build a real
    npm layout in a temp dir, drive it with { platform: 'win32' }, and
    assert the payload reaches the recorded argv byte-for-byte while the
    sibling .cmd/.ps1 are resolved past, never executed.
  - the coverage is added to plugins/ruflo-core/scripts/test-hooks.mjs,
    whose "Plugin hooks smoke" job already runs on windows-latest. The
    existing cases there go through RUFLO_HOOK_CLI_OVERRIDE and so never
    reach the global-shim branch; these call invokeHook() directly, so the
    fallback is executed against a real cmd.exe.

Tests: 11 vitest cases and 6 new harness cases, each mutation-checked —
every one fails when the code it covers is reverted. test-hooks.mjs goes
19/24 to 25/30 against the recorder fixture; those 5 failures are
pre-existing on main (the fixture does not echo argv) and pass in CI
against the real built CLI.

Windows result: the windows-latest leg ran green, 32/32, including the
escaped fallback executed against a real cmd.exe. The %VAR% probe was
included because carets are not a reliable escape for % and quoting does
not suppress percent expansion, so expansion on the shim's second parse
looked plausible; the runner measured otherwise and the value arrives
literal, with no redirection performed. The probe stays as a regression
guard and keeps reporting the observed value rather than asserting one.

Out of scope: nine sites in plugins/ruflo-metaharness/scripts spawn with
`shell: process.platform === 'win32'` and a dynamic argv element —
_darwin.mjs:86/129 spread the caller's own ...argv, oia-audit.mjs:105 passes
JSON.stringify(payload), and audit-list/audit-trend/similarity pass
`--key <key>`. Same shape, different plugin, separate change.

Two neighbours that look like the same problem and are not: the ten `shell:`
flags across ruflo-cost-tracker are inert, because spawnNpxSync() discards
the option and forces shell:false; and mcp-launch.cjs already prefers a
resolved local bin with shell:false and only falls back to npx.cmd with
constant args.
2026-09-16 14:36:26 +00:00
..

🚀 Claude Flow Plugin - Complete Enterprise AI Agent Orchestration

Version License Claude Code

Enterprise-grade AI agent orchestration plugin with 150+ commands, 74+ specialized agents, SPARC methodology, swarm coordination, GitHub integration, and neural training capabilities


📋 Table of Contents


🌟 Overview

Claude Flow is the most comprehensive Claude Code plugin for enterprise AI agent orchestration. It provides a complete ecosystem for:

  • Multi-Agent Coordination: 74+ specialized agents with swarm intelligence
  • SPARC Methodology: Systematic development with 18 specialized modes
  • GitHub Automation: 14+ tools for complete repository workflow automation
  • Neural Training: 27+ models with WASM acceleration
  • 150+ Commands: Complete slash command library for all workflows
  • MCP Integration: 110+ tools across 3 MCP servers

✨ Features

🐝 Swarm Coordination

  • 4 Topologies: Hierarchical, Mesh, Ring, Star
  • Auto-Spawning: Intelligent agent creation based on task complexity
  • Auto-Optimization: Dynamic topology adjustment for performance
  • 100 Max Agents: Scale to handle enterprise workloads
  • Cross-Session Memory: Persistent context and learnings

🎯 SPARC Methodology

  • Specification: Requirements analysis and planning
  • Pseudocode: Algorithm design and logic flow
  • Architecture: System design and component structure
  • Refinement: TDD and iterative improvement
  • Code: Implementation and optimization
  • 18 Specialized Modes: Complete development lifecycle coverage

🐙 GitHub Integration

  • PR Management: Automated pull request workflows
  • Code Review Swarms: Multi-agent code analysis
  • Issue Tracking: Intelligent issue triage and assignment
  • Release Automation: Coordinated multi-package releases
  • Workflow Automation: Custom GitHub Actions integration
  • Multi-Repo Coordination: Cross-repository synchronization

🧠 Neural Training

  • 27+ Models: Pre-trained patterns for common tasks
  • WASM Acceleration: 2.8-4.4x speed improvement
  • SIMD Optimization: Advanced vector processing
  • Pattern Learning: Self-improving agent behaviors
  • Context Persistence: Cross-session learning retention

🎨 74+ Specialized Agents

Core Development (5)

  • coder - Code implementation specialist
  • planner - Strategic planning and roadmaps
  • researcher - Information gathering and analysis
  • reviewer - Code quality and security review
  • tester - Comprehensive test creation

Swarm Coordination (5)

  • hierarchical-coordinator - Queen-led command structure
  • mesh-coordinator - Peer-to-peer coordination
  • adaptive-coordinator - Dynamic topology management
  • collective-intelligence-coordinator - Distributed decision-making
  • swarm-memory-manager - Cross-agent memory coordination

Consensus & Fault Tolerance (7)

  • byzantine-coordinator - Byzantine fault tolerance
  • raft-manager - Raft consensus protocol
  • gossip-coordinator - Gossip-based consensus
  • crdt-synchronizer - Conflict-free data replication
  • quorum-manager - Dynamic quorum management
  • security-manager - Comprehensive security protocols
  • performance-benchmarker - Consensus performance testing

GitHub Automation (13)

  • pr-manager - Pull request coordination
  • code-review-swarm - Multi-agent code reviews
  • issue-tracker - Issue management and triage
  • release-manager - Release coordination
  • workflow-automation - GitHub Actions management
  • repo-architect - Repository structure optimization
  • multi-repo-swarm - Cross-repository coordination
  • sync-coordinator - Version alignment across repos
  • And 5 more specialized GitHub agents...

Specialized Development (8)

  • backend-dev - Backend API development
  • mobile-dev - React Native mobile development
  • ml-developer - Machine learning workflows
  • cicd-engineer - CI/CD pipeline creation
  • api-docs - OpenAPI/Swagger documentation
  • system-architect - System design and architecture
  • code-analyzer - Advanced code quality analysis
  • base-template-generator - Boilerplate generation

SPARC Methodology (4)

  • specification - Requirements analysis
  • pseudocode - Algorithm design
  • architecture - System architecture
  • refinement - Iterative improvement

And 32 more specialized agents!

📦 150+ Commands

Coordination (6)

  • /coordination-swarm-init - Initialize swarm with topology
  • /coordination-agent-spawn - Create specialized agents
  • /coordination-task-orchestrate - Coordinate task execution
  • /coordination-spawn - Quick agent spawning
  • /coordination-orchestrate - Advanced orchestration
  • /coordination-init - Setup coordination environment

SPARC Methodology (18)

  • /sparc-modes - List all SPARC modes
  • /sparc-coder - Clean code implementation
  • /sparc-tdd - Test-driven development
  • /sparc-architect - Architecture design
  • /sparc-reviewer - Code review mode
  • /sparc-tester - Test creation mode
  • /sparc-analyzer - Code analysis
  • /sparc-researcher - Research mode
  • /sparc-optimizer - Performance optimization
  • /sparc-debugger - Debugging assistance
  • /sparc-designer - UI/UX design mode
  • /sparc-documenter - Documentation creation
  • /sparc-innovator - Innovation and R&D
  • /sparc-orchestrator - Workflow orchestration
  • /sparc-batch-executor - Batch operations
  • /sparc-memory-manager - Memory management
  • /sparc-workflow-manager - Workflow management
  • /sparc-swarm-coordinator - Swarm coordination

GitHub Integration (18)

  • /github-code-review - Automated code reviews
  • /github-code-review-swarm - Multi-agent reviews
  • /github-pr-manager - PR lifecycle management
  • /github-pr-enhance - PR enhancement automation
  • /github-issue-tracker - Issue tracking
  • /github-issue-triage - Intelligent issue triage
  • /github-repo-analyze - Repository analysis
  • /github-repo-architect - Repo structure optimization
  • /github-release-manager - Release coordination
  • /github-release-swarm - Multi-package releases
  • /github-workflow-automation - GitHub Actions automation
  • /github-swarm-pr - PR swarm management
  • /github-swarm-issue - Issue swarm coordination
  • /github-multi-repo-swarm - Cross-repo coordination
  • /github-sync-coordinator - Version synchronization
  • /github-project-board-sync - Project board integration
  • /github-modes - GitHub integration modes
  • /github-swarm - GitHub swarm orchestration

Hive Mind (11)

  • /hive-mind - Initialize hive mind coordination
  • /hive-mind-init - Setup hive mind topology
  • /hive-mind-spawn - Spawn hive agents
  • /hive-mind-status - Check hive status
  • /hive-mind-consensus - Consensus protocols
  • /hive-mind-memory - Shared memory management
  • /hive-mind-metrics - Performance metrics
  • /hive-mind-sessions - Session management
  • /hive-mind-resume - Resume hive sessions
  • /hive-mind-stop - Stop hive coordination
  • /hive-mind-wizard - Guided setup wizard

Memory Management (5)

  • /memory-usage - Memory storage and retrieval
  • /memory-persist - Cross-session persistence
  • /memory-search - Pattern-based search
  • /memory-neural - Neural memory integration

Monitoring (5)

  • /monitoring-status - System status overview
  • /monitoring-agents - Agent status monitoring
  • /monitoring-agent-metrics - Performance metrics
  • /monitoring-swarm-monitor - Real-time swarm monitoring
  • /monitoring-real-time-view - Live dashboard

Optimization (5)

  • /optimization-topology-optimize - Auto-optimize topology
  • /optimization-auto-topology - Automatic topology selection
  • /optimization-parallel-execution - Parallel task execution
  • /optimization-parallel-execute - Execute tasks in parallel
  • /optimization-cache-manage - Cache management

Analysis (5)

  • /analysis-performance-report - Performance reports
  • /analysis-performance-bottlenecks - Bottleneck detection
  • /analysis-bottleneck-detect - Real-time bottleneck analysis
  • /analysis-token-usage - Token consumption analysis
  • /analysis-token-efficiency - Token optimization

Automation (6)

  • /automation-smart-spawn - Intelligent agent spawning
  • /automation-smart-agents - Auto-agent selection
  • /automation-auto-agent - Automated agent management
  • /automation-self-healing - Self-healing workflows
  • /automation-session-memory - Session persistence
  • /automation-workflow-select - Workflow selection

Hooks (7)

  • /hooks-setup - Configure hooks system
  • /hooks-overview - Hooks documentation
  • /hooks-pre-task - Pre-task hook setup
  • /hooks-post-task - Post-task hook setup
  • /hooks-pre-edit - Pre-edit hook setup
  • /hooks-post-edit - Post-edit hook setup
  • /hooks-session-end - Session end hook setup

Swarm Management (15)

  • /swarm - Main swarm command
  • /swarm-init - Initialize swarm
  • /swarm-spawn - Spawn swarm agents
  • /swarm-status - Swarm status
  • /swarm-monitor - Real-time monitoring
  • /swarm-modes - Available swarm modes
  • /swarm-strategies - Execution strategies
  • /swarm-background - Background swarm execution
  • /swarm-analysis - Swarm analysis workflows
  • /swarm-research - Research swarms
  • /swarm-development - Development swarms
  • /swarm-testing - Testing swarms
  • /swarm-maintenance - Maintenance swarms
  • /swarm-optimization - Optimization swarms
  • /swarm-examples - Swarm examples

Workflows (5)

  • /workflows-create - Create custom workflows
  • /workflows-execute - Execute workflows
  • /workflows-export - Export workflow definitions
  • /workflows-development - Development workflows
  • /workflows-research - Research workflows

Neural Training (5)

  • /training-neural-train - Train neural patterns
  • /training-neural-patterns - Pattern management
  • /training-pattern-learn - Pattern learning
  • /training-model-update - Model updates
  • /training-specialization - Agent specialization

Flow Nexus (9)

  • /flow-nexus-swarm - Cloud swarm orchestration
  • /flow-nexus-workflow - Event-driven workflows
  • /flow-nexus-neural-network - Distributed neural training
  • /flow-nexus-sandbox - E2B sandbox management
  • /flow-nexus-app-store - Application marketplace
  • /flow-nexus-challenges - Coding challenges
  • /flow-nexus-payments - Credit management
  • /flow-nexus-user-tools - User management
  • /flow-nexus-login - Authentication

And 50+ more commands!


🚀 Quick Start

1. Install Claude Code Plugin

In Claude Code:

/plugin add ruvnet/claude-flow

Or from local directory:

git clone https://github.com/ruvnet/claude-flow.git
cd claude-flow

Then in Claude Code:

/plugin add .

2. Restart Claude Code

/restart

3. Configure MCP Servers (Optional)

# Add MCP servers to Claude Code
claude mcp add claude-flow npx claude-flow@alpha mcp start
claude mcp add ruv-swarm npx ruv-swarm mcp start  # Optional
claude mcp add flow-nexus npx flow-nexus@latest mcp start  # Optional

4. Verify Installation

# Check plugin status
claude plugin list

# Test a command
# In Claude Code, type:
/coordination-swarm-init

📦 Installation

Prerequisites

  • Claude Code CLI >= 2.0.0
  • Node.js >= 20.0.0
  • Git (for GitHub integration features)
  • Read/write permissions in project directory

In Claude Code:

/plugin add ruvnet/claude-flow
/restart

Method 2: Local Installation

# Clone the repository
git clone https://github.com/ruvnet/claude-flow.git
cd claude-flow/claude-plugin

# Run installation script
bash scripts/install.sh

# Or copy manually
cp -r commands ~/.claude/commands/
cp -r agents ~/.claude/agents/

Method 3: NPX (One-Time Setup)

# Run setup via npx
npx claude-flow@alpha init --plugin

# This will:
# 1. Create .claude directory
# 2. Copy all commands and agents
# 3. Configure MCP servers
# 4. Setup hooks

🏗️ Components

Directory Structure

claude-flow/
├── .claude-plugin/
│   ├── plugin.json          # Plugin metadata
│   ├── README.md            # This file
│   └── ...
├── commands/                 # 150+ slash commands
│   ├── coordination/         # Swarm coordination commands
│   ├── sparc/                # SPARC methodology commands
│   ├── github/               # GitHub integration commands
│   ├── hive-mind/            # Hive mind commands
│   ├── hooks/                # Hooks configuration commands
│   ├── memory/               # Memory management commands
│   ├── monitoring/           # Monitoring commands
│   ├── optimization/         # Optimization commands
│   ├── analysis/             # Analysis commands
│   ├── automation/           # Automation commands
│   ├── swarm/                # Swarm management commands
│   ├── workflows/            # Workflow commands
│   ├── training/             # Neural training commands
│   ├── flow-nexus/           # Flow Nexus integration
│   └── ...                   # And more!
├── agents/                   # 74+ specialized agents
│   ├── core/                 # Core development agents
│   ├── consensus/            # Consensus protocol agents
│   ├── github/               # GitHub automation agents
│   ├── swarm/                # Swarm coordination agents
│   ├── hive-mind/            # Hive mind agents
│   ├── sparc/                # SPARC methodology agents
│   ├── optimization/         # Optimization agents
│   ├── specialized/          # Domain-specific agents
│   ├── templates/            # Template agents
│   ├── testing/              # Testing agents
│   └── ...                   # And more!
├── hooks/                    # Hook scripts
│   ├── pre-tool-use.sh
│   ├── post-tool-use.sh
│   ├── pre-task.sh
│   ├── post-task.sh
│   ├── session-start.sh
│   └── session-end.sh
├── scripts/                  # Installation and setup scripts
│   ├── install.sh
│   ├── setup-mcp.sh
│   ├── verify.sh
│   └── uninstall.sh
└── docs/                     # Documentation
    ├── QUICKSTART.md
    ├── USER_GUIDE.md
    ├── API_REFERENCE.md
    ├── EXAMPLES.md
    └── TROUBLESHOOTING.md

💡 Usage

Basic Swarm Coordination

# Initialize a hierarchical swarm
/coordination-swarm-init

# Spawn specialized agents
/coordination-agent-spawn

# Orchestrate a complex task
/coordination-task-orchestrate "Build REST API with authentication"

SPARC Development Workflow

# Start with specification
/sparc-modes specification "User authentication system"

# Design architecture
/sparc-architect

# Implement with TDD
/sparc-tdd "Implement JWT authentication"

# Code review
/sparc-reviewer

# Optimize performance
/sparc-optimizer

GitHub Automation

# Analyze repository
/github-repo-analyze

# Create PR with automated review
/github-pr-manager

# Multi-agent code review
/github-code-review-swarm

# Coordinate release across repos
/github-multi-repo-swarm

Hive Mind Coordination

# Initialize hive mind
/hive-mind-init

# Spawn hive agents with consensus
/hive-mind-spawn

# Check consensus status
/hive-mind-consensus

# View shared memory
/hive-mind-memory

🔌 MCP Integration

Claude Flow integrates with 3 MCP servers providing 110+ tools:

Claude Flow MCP (Required)

{
  "mcpServers": {
    "claude-flow": {
      "command": "npx",
      "args": ["claude-flow@alpha", "mcp", "start"]
    }
  }
}

Tools: 40+ orchestration tools

  • Swarm initialization and management
  • Agent spawning and coordination
  • Task orchestration
  • Memory management
  • Neural training
  • Performance monitoring

ruv-swarm MCP (Optional)

{
  "mcpServers": {
    "ruv-swarm": {
      "command": "npx",
      "args": ["ruv-swarm", "mcp", "start"]
    }
  }
}

Tools: Enhanced coordination features

  • WASM acceleration (2.8-4.4x speed)
  • SIMD optimization
  • Advanced topology management
  • Byzantine fault tolerance

Flow Nexus MCP (Optional - Requires Auth)

{
  "mcpServers": {
    "flow-nexus": {
      "command": "npx",
      "args": ["flow-nexus@latest", "mcp", "start"]
    }
  }
}

Tools: 70+ cloud features

  • E2B sandbox execution
  • Distributed neural training
  • Event-driven workflows
  • Application marketplace
  • Real-time collaboration

📚 Examples

Example 1: Full-Stack Development with Swarm

# Initialize hierarchical swarm
/coordination-swarm-init

# The swarm automatically spawns:
# - backend-dev agent
# - coder agent for frontend
# - tester agent
# - reviewer agent

# Orchestrate the full-stack build
/coordination-task-orchestrate "Build a todo app with React frontend and Express backend"

# Monitor progress
/monitoring-swarm-monitor

# Get performance metrics
/analysis-performance-report

Example 2: SPARC TDD Workflow

# Start with specification
/sparc-modes specification "Shopping cart with inventory management"

# Generate pseudocode
/sparc-modes pseudocode

# Design architecture
/sparc-architect

# TDD implementation
/sparc-tdd

# Automated review
/sparc-reviewer

# Performance optimization
/sparc-optimizer

Example 3: GitHub PR Automation

# Analyze current PR
/github-pr-manager

# Multi-agent code review
/github-code-review-swarm

# Auto-fix issues
/github-pr-enhance

# Sync across repositories
/github-sync-coordinator

# Prepare release
/github-release-manager

📖 Documentation


🤝 Support


📊 Performance

  • 84.8% SWE-Bench solve rate
  • 32.3% token reduction vs. sequential execution
  • 2.8-4.4x speed improvement with WASM acceleration
  • 27+ neural models for pattern recognition
  • 100 max concurrent agents

🔧 Advanced Configuration

Custom Swarm Topology

{
  "swarmCoordination": {
    "topology": "mesh",
    "maxAgents": 50,
    "autoSpawn": true,
    "autoOptimize": true
  }
}

Enable Neural Training

{
  "neuralTraining": {
    "enabled": true,
    "wasmAcceleration": true,
    "simdOptimization": true
  }
}

Configure Hooks

{
  "hooks": {
    "PreToolUse": { "enabled": true },
    "PostToolUse": { "enabled": true },
    "SessionEnd": { "enabled": true }
  }
}

📝 License

MIT License - see LICENSE file for details


🌟 Star History

Star History Chart


Made with ❤️ by rUv

Enterprise AI Agent Orchestration for Claude Code