multi-agent

Orchestrate parallel worker subagents in isolated worktrees to explore task variants.

Updated Apr 10, 2026
One-click install
npx skills add https://github.com/erichugy/agentic.skills --skill multi-agent-erichugy
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: multi-agent
Source: https://github.com/erichugy/agentic.skills/tree/main/multi-agent
Command: npx skills add https://github.com/erichugy/agentic.skills --skill multi-agent-erichugy

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve?

Breaks a complex task into parallel variants and manages isolated worker subagents so teams can explore multiple approaches at once without manual branching, coordination, or ad-hoc reviews.

Core Features & Use Cases

  • Parallel variant orchestration: Plan and launch multiple worker subagents running in isolated worktrees to implement different design or implementation directions concurrently.
  • Wave-based dependency management: Define dependency chains and waves so dependent workers branch off prerequisite branches and only start when prerequisites succeed.
  • Iterative review and hardening loop: Automatically launch reviewers with a production hardening checklist, collect issues, apply fixes, and repeat until all reviewers pass.
  • Branching and PR chains: Produce consistent branch names for each variant and generate PR chains in reverse-wave order for dependent changes.
  • Error handling and reporting: Mark blocked downstream work on failure, offer retry/skip options, and summarize outcomes for user decision-making.

Quick Start

Launch a plan that creates three variants, run workers in parallel across isolated worktrees, iterate reviewers until all pass, and present a summarized comparison with recommended next steps.

Frequently Asked Questions about multi-agent

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I run parallel code variants in isolated worktrees to compare approaches?

Parallel variant orchestration launches multiple worker subagents in isolated git worktrees to implement different design directions concurrently. Workers operate independently, allowing teams to explore multiple approaches simultaneously without manual branching or coordination.

What is wave-based dependency management for parallel worker agents?

Wave-based dependency management defines dependency chains so dependent workers branch off prerequisite branches and only start when prerequisites succeed. This ensures ordered execution across parallel variants while maintaining correct branch lineage for downstream work.

How does automated iterative code review and hardening work?

Iterative review and hardening automatically launches reviewers with a production hardening checklist, collects issues, applies fixes, and repeats until all reviewers pass. This loop runs without manual intervention, ensuring variants meet quality standards before comparison.

Can I create PR chains for dependent branches in reverse-wave order?

Yes, the orchestration produces consistent branch names for each variant and generates PR chains in reverse-wave order for dependent changes. This ensures that pull requests are stacked correctly when downstream work depends on upstream variants.

How are failures and blocked downstream work handled during parallel orchestration?

Error handling marks blocked downstream work on failure and offers retry or skip options. The system summarizes outcomes for user decision-making, allowing teams to decide whether to reattempt failed variants or proceed with successful alternatives.

When should I use multi-agent parallel workers instead of sequential implementation?

Use parallel workers when a complex task has multiple viable design or implementation directions that should be explored concurrently. It breaks tasks into independent variants running in isolated worktrees, eliminating manual branching, coordination, and ad-hoc reviews for sequential workflows.