Automation Tools | AI General

🎭 Browser Automation (Playwright CLI)

4-layer architecture for scalable, production-grade browser automation.

Architecture Layers

1 Skills - High-level workflows (visual regression, form testing)

2 Agents - Specialized browser agents (Screenshot, FormTest, Navigate)

3 Justfile - Task runner (just browser-screenshot, just test-form)

4 Playwright CLI - Core execution engine

just browser-screenshot https://example.com
just browser-test-form login-flow.spec.ts
just browser-navigate-multi page1.html page2.html page3.html

⚡ Workflow Orchestration

Agent coordination and state management for complex multi-step workflows.

Parallel Execution

Run multiple agents simultaneously

Agent Specialization

Custom agents for specific tasks

State Management

Persist workflow state across sessions

Error Recovery

Automatic retry and fallback strategies

// Multi-agent research workflow
const agents = [
  { type: 'YouTubeResearch', query: 'AI research' },
  { type: 'ArXivSearch', query: 'multi-agent' },
  { type: 'FabricExtract', pattern: 'extract_wisdom' }
];
await orchestrateParallel(agents);

🤝 Task Delegation

Six patterns for delegating work across specialized agents.

Built-in Agents

Engineer, Architect, Algorithm via Task tool

Worktree-Isolated

Parallel file edits without conflicts

Background Agents

Non-blocking execution (run_in_background: true)

Custom Agents

ComposeAgent for specialized tasks

Agent Teams

Multi-turn peer coordination (TeamCreate)

Parallel Dispatch

N identical operations fan-out

// Built-in agent delegation
Agent({ 
  subagent_type: "Engineer", 
  model: "sonnet",
  task: "Implement authentication module"
});

// Agent team coordination
TeamCreate({ members: ["Engineer", "QATester", "SecurityAuditor"] });
TaskCreate({ team_id, description: "Build user login system" });