Multi Agent Patterns

When you're hitting context window limits or need to run parallel tasks with different tool sets, this skill gives you architectural patterns for splitting work across multiple Claude instances. The key insight here is smart: treat sub-agents as context isolation boundaries, not as role-playing exercises. It covers three coordination patterns based on actual needs rather than organizational metaphors. With 162 installs and solid audit results, it's seeing real use. Worth noting that the description is sparse, but the core premise is sound. If you're building production systems that need to juggle multiple domains simultaneously or scale beyond what one context window can handle, the patterns here will save you from reinventing coordination overhead.

Install to Claude Code

npx -y skills add flora131/atomic --skill multi-agent-patterns --agent claude-code