How does the Subagent Teams skill prevent performance degradation?

It isolates heavy tasks like codebase searching and testing into separate subagent context windows, keeping the main orchestrator's context lean and preventing the 'auto-compact' feature from truncating important information.

Can I run multiple subtasks at the same time?

Yes, the skill identifies independent tasks and launches them in a single message via parallel tool calls, significantly reducing the total time required for complex workflows.

Which models are used for the subagents?

The skill intelligently selects models based on complexity: Haiku for simple searches, Sonnet for analysis and design, and Opus for critical reasoning or multi-step implementations.

When should I avoid using the Subagent Teams skill?

You should not use this skill for single-step trivial tasks or tasks with strict sequential dependencies where one step must finish before the next can even be planned.

Subagent Teams Orchestrator

Name: Subagent Teams Orchestrator
Author: alijilani-dev

byalijilani-dev

0•

生産性とワークフロー

Orchestrates complex development tasks by delegating work to parallel subagent teams to maintain optimal performance and context integrity.

The Subagent Teams skill empowers Claude to manage large-scale projects by breaking down complex requirements into independent, manageable subtasks. By delegating exploration, testing, and implementation to specialized subagents operating in separate context windows, this skill prevents the main session's context from becoming bloated. This results in faster execution through parallelism and higher quality outputs by avoiding the 'auto-compact' feature that often degrades reasoning in long-running sessions.

主な機能

01Parallel task decomposition and execution for faster workflows

02Context isolation to prevent auto-compact and maintain reasoning quality

030 GitHub stars

04Automated result synthesis to merge findings into a single action plan

05Intelligent model selection (Opus, Sonnet, Haiku) based on subtask complexity

06Specialized agent roles for exploration, planning, and implementation

ユースケース

01Exploring large codebases and analyzing patterns without filling the main context window

02Simultaneous implementation of multiple independent features or modules

03Running complex test suites and error analysis in parallel with code development

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add alijilani-dev/claude subagent-teams

For use in Claude.ai and ChatGPT

Download Skill