AgentSkillsCN

antfarm

为OpenClaw构建多智能体工作流。在打造受益于计划→实施→验证→测试→评审循环的复杂功能时使用此功能。

SKILL.md
--- frontmatter
name: antfarm
description: Multi-agent workflows for OpenClaw. Use when building complex features that benefit from plan → implement → verify → test → review cycles.

Antfarm + VibeKit

Use Antfarm for sophisticated multi-agent development, VibeKit for deployment.

When to Use

  • •Complex features needing multiple implementation steps
  • •Security audits with automated fixes
  • •Bug fixes requiring investigation and regression tests
  • •Any task benefiting from verification gates

Setup

bash
# Install antfarm (requires OpenClaw locally)
npx antfarm install

# Run a feature workflow
antfarm workflow run feature-dev "Add user authentication with OAuth"

# Check status
antfarm workflow status "OAuth"

Workflows

WorkflowAgentsUse Case
feature-dev7Feature requests → tested PR
security-audit7Vulnerability scan → fix PR
bug-fix6Bug report → fix with regression test

With VibeKit

Option 1: Antfarm develops, VibeKit deploys

bash
# Antfarm creates the PR
antfarm workflow run feature-dev "Add Stripe checkout"

# VibeKit deploys when PR merges
curl -X POST https://vibekit.bot/api/v1/task \
  -H "Authorization: Bearer $VIBEKIT_KEY" \
  -d '{"prompt": "Deploy latest from main", "repo": "user/app"}'

Option 2: VibeKit for quick iterations, Antfarm for complex features

  • •Simple UI changes → VibeKit (instant)
  • •New auth system → Antfarm (thorough)

How It Works

code
plan → setup → implement → verify → test → PR → review
  │       │         │         │        │      │       │
  └───────┴─────────┴─────────┴────────┴──────┴───────┘
              Each agent = fresh context
              Failed steps retry automatically
              Nothing ships without review

Requirements

  • •Node.js >= 22
  • •OpenClaw running locally
  • •gh CLI for PR creation

Resources