How do I start agents after initialization?

After running /hub:init to create the session, use the /hub:spawn command to launch the agents and begin the work.

What happens during the baseline capture phase?

If an evaluation command is provided, the skill runs that command against your current codebase to establish a 'starting point' metric, allowing you to track improvements made by the agents.

Can I use this skill without an evaluation command?

Yes, if no evaluation command is provided, the session defaults to LLM judge mode, which is ideal for research, drafting, and creative tasks where objective metrics are not applicable.

Where are the session details stored?

All configurations, session IDs, and state information are stored locally in the .agenthub/ directory within your project's root folder.

What is an AgentHub session?

An AgentHub session is a structured collaboration environment where multiple AI agents work together on a specific task defined by the user within Claude Code.

AgentHub Session Initializer

Name: AgentHub Session Initializer
Author: micsapp

bymicsapp

0•

コラボレーションツール

Initializes multi-agent collaboration sessions by configuring tasks, agent counts, and automated evaluation criteria.

The AgentHub Session Initializer is a specialized skill for Claude Code designed to orchestrate complex multi-agent workflows. It automates the setup of parallel AI agent sessions by creating the necessary directory structures, generating session IDs, and defining success metrics. Whether you are performing objective code optimizations that require benchmarking or creative content tasks that utilize LLM-based judging, this skill ensures that your multi-agent environment is properly configured and baselined. By providing a structured framework for task delegation and performance tracking, it allows developers to scale their productivity through coordinated AI collaboration.

主な機能

01Support for both interactive and argument-driven command modes

02Automated session directory structure and configuration generation

03Configurable agent counts for parallel task processing

040 GitHub stars

05Flexible evaluation modes including command-line metrics and LLM judging

06Built-in baseline metric capture for performance-driven tasks

ユースケース

01Orchestrating complex refactoring jobs across multiple parallel AI agents

02Setting up code optimization tasks with automated performance benchmarking

03Launching multi-agent brainstorming or creative content generation sessions

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add micsapp/micstec-skills skills-init

For use in Claude.ai and ChatGPT