@claude-flow/hooks/src/workers/
├── index.ts           # WorkerManager, all worker implementations
├── mcp-tools.ts       # MCP tool definitions for workers
├── session-hook.ts    # Claude Code session integration
└── __tests__/         # Comprehensive test suite
    └── workers.test.ts

class WorkerManager extends EventEmitter {
  // Core methods
  register(name: string, handler: WorkerHandler): void;
  async initialize(): Promise<void>;
  async start(options?: StartOptions): Promise<void>;
  async stop(): Promise<void>;
  async runWorker(name: string): Promise<WorkerResult>;
  async runAll(concurrency?: number): Promise<WorkerResult[]>;

  // Persistence
  async loadState(): Promise<boolean>;
  async saveState(): Promise<void>;

  // Alerts
  setThresholds(worker: string, thresholds: AlertThreshold[]): void;
  getAlerts(limit?: number): WorkerAlert[];
  clearAlerts(): void;

  // History
  getHistory(worker?: string, limit?: number): HistoricalMetric[];

  // Statusline
  getStatuslineData(): StatuslineData;
  getStatuslineString(): string;
  async exportStatusline(): Promise<void>;
}

const DEFAULT_THRESHOLDS = {
  health: [
    { metric: 'memory.usedPct', warning: 80, critical: 95, comparison: 'gt' },
    { metric: 'disk.usedPct', warning: 85, critical: 95, comparison: 'gt' },
  ],
  security: [
    { metric: 'secrets', warning: 1, critical: 5, comparison: 'gt' },
    { metric: 'vulnerabilities', warning: 10, critical: 50, comparison: 'gt' },
  ],
  adr: [
    { metric: 'compliance', warning: 70, critical: 50, comparison: 'lt' },
  ],
};

// Auto-start on session begin
export async function onSessionStart(config?: SessionHookConfig): Promise<SessionHookResult>;

// Clean shutdown on session end
export async function onSessionEnd(manager: WorkerManager): Promise<void>;

Tests:       45 passed (45)
Duration:    1.2s

Coverage:
- WorkerManager: Unit tests
- Alert System: Threshold tests
- Historical Metrics: Recording tests
- Statusline: Export tests
- Persistence: Save/load tests
- Security: Boundary tests
- Built-in Workers: Integration tests

import { createWorkerManager } from '@claude-flow/hooks';

const manager = createWorkerManager('/path/to/project');
await manager.initialize();
await manager.start();

// Run specific worker
const result = await manager.runWorker('security');
console.log(result.data);

// Get alerts
const alerts = manager.getAlerts(10);

// Get statusline
const statusline = manager.getStatuslineString();
// "👷0/10 │ 🟢15% │ 🛡️0 │ 📋71% │ 🏗️13% │ ⚡1.0x"

import { createWorkerToolHandler, workerMCPTools } from '@claude-flow/hooks';

// Register tools with MCP server
const handler = createWorkerToolHandler(manager);

// Handle tool call
const result = await handler('worker/run', { worker: 'health' });

import { onSessionStart, formatSessionStartOutput } from '@claude-flow/hooks';

const result = await onSessionStart({
  projectRoot: '/path/to/project',
  autoStart: true,
  runInitialScan: true,
  workers: ['health', 'security', 'git'],
});

console.log(formatSessionStartOutput(result));
// [Workers] System initialized
//   ✓ Health: healthy
//   ✓ Security: clean (0 issues)
//   ├─ Branch: v3
//   └─ Uncommitted: 5
// [Workers] Background scheduling started

# List all available workers
claude-flow hooks worker list

# Detect triggers from prompt text (<5ms target)
claude-flow hooks worker detect --prompt "optimize performance"

# Auto-dispatch when triggers match (confidence ≥0.6)
claude-flow hooks worker detect --prompt "deep dive" --auto-dispatch --min-confidence 0.6

# Manually dispatch a worker
claude-flow hooks worker dispatch --trigger refactor --context "auth module"

# Check worker status
claude-flow hooks worker status

# Cancel a running worker
claude-flow hooks worker cancel --id worker_refactor_1_abc123

{
  "hooks": {
    "UserPromptSubmit": [{
      "matcher": ".*",
      "hooks": [{
        "type": "command",
        "timeout": 6000,
        "command": "claude-flow hooks worker detect --prompt \"$USER_PROMPT\" --auto-dispatch --min-confidence 0.6"
      }]
    }]
  }
}

# Start the daemon (runs workers on intervals)
npx claude-flow@v3alpha daemon start
npx claude-flow@v3alpha daemon start --quiet  # Run once and exit

# Stop the daemon
npx claude-flow@v3alpha daemon stop

# Check status and worker history
npx claude-flow@v3alpha daemon status

# Manually trigger a worker
npx claude-flow@v3alpha daemon trigger <worker>
npx claude-flow@v3alpha daemon trigger map --force

# Enable/disable workers
npx claude-flow@v3alpha daemon enable map audit optimize
npx claude-flow@v3alpha daemon enable --all

.claude-flow/metrics/
├── codebase-map.json      # map worker output
├── security-audit.json    # audit worker output
├── performance.json       # optimize worker output
├── consolidation.json     # consolidate worker output
├── test-gaps.json         # testgaps worker output
├── agent-metrics.json     # Agent performance data
└── task-metrics.json      # Task execution data

interface DaemonState {
  workers: {
    [key: string]: {
      enabled: boolean;
      runCount: number;
      successCount: number;
      failureCount: number;
      lastRun?: Date;
      lastError?: string;
    };
  };
  pid?: number;
  startedAt?: string;
}

// Auto-start on SessionStart hook
hooks.SessionStart = [{
  hooks: [{
    type: 'command',
    command: 'npx claude-flow@v3alpha daemon start --quiet 2>/dev/null || true',
    timeout: 5000,
    continueOnError: true,
  }]
}];

{
  "name": "claude-flow",
  "bin": {
    "claude-flow": "./v3/@claude-flow/cli/bin/cli.js"
  },
  "publishConfig": {
    "access": "public",
    "tag": "v3alpha"
  }
}