When should I use a sub-agent instead of a skill?

Use a sub-agent when task complexity is high (score 7-8). For simpler, repetitive tasks, a standard skill is usually the more efficient approach.

Does it support different agent types?

Yes, it provides templates for various levels of complexity, including autonomous agents, read-only agents, and agents with specific skills or hooks.

Can I validate my agent files before deployment?

Yes, the skill includes built-in scripts to validate frontmatter and agent file structure to ensure they meet the required specifications.

What is a Claude sub-agent?

A sub-agent is a specialized instance of Claude designed to handle specific tasks with its own unique set of tools, instructions, and context boundaries.

How does this skill handle security and permissions?

It implements a 'least privilege' tool access strategy, ensuring sub-agents only have access to the specific tools required for their defined role.

Sub-Agent Implementation

Name: Sub-Agent Implementation
Author: yu-iskw

byyu-iskw

0•

生产力与工作流

Automates the creation, validation, and configuration of specialized Claude sub-agents with precise tool permissions and delegation boundaries.

This skill streamlines the development of Claude sub-agents by providing a structured framework for defining roles, setting decision-making authority, and configuring tool access. It helps developers determine when a sub-agent is necessary based on task complexity, ensures adherence to Anthropic's best practices, and includes built-in validation scripts for frontmatter and file structure. By utilizing pre-defined templates for autonomous, read-only, or hook-enabled agents, it reduces the friction in building robust, multi-agent workflows within the Claude ecosystem.

主要功能

01Evaluation planning tools to identify success and failure modes across test scenarios

020 GitHub stars

03Pre-configured templates for autonomous, read-only, and hook-enabled agents

04Complexity scoring to determine when sub-agents are required over standard skills

05Context management strategies for retrieval, summarization, and state preservation

06Automated validation for agent markdown frontmatter and file structure

使用场景

01Building specialized read-only agents for codebase analysis and documentation

02Implementing autonomous agents with strictly defined tool permissions and boundaries

03Architecting multi-agent systems for complex software engineering projects

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add yu-iskw/claude-plugin-builder implement-sub-agents

For use in Claude.ai and ChatGPT

Download Skill