How does this skill handle sub-agent security?

It implements a least-privilege tool access strategy, allowing you to define explicit permission boundaries so sub-agents only access the specific tools required for their role.

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

Sub-agents should be used for high-complexity tasks (scoring 7-8 on the complexity framework) that require distinct role boundaries or specialized toolsets. For simpler, repeatable utilities, a standard skill is preferred.

Can I validate my sub-agent configuration before deployment?

Yes, the skill includes dedicated validation tools and frontmatter checkers to ensure agent files follow correct structures and delegation rules.

What types of agent templates are included?

The skill provides templates for level-3 autonomous agents, base sub-agents, read-only agents, and agents integrated with specific skills or hooks.

Claude Sub-Agent Implementation

Name: Claude Sub-Agent Implementation
Author: yu-iskw

byyu-iskw

0•

開発者ツール

Architects and validates specialized Claude sub-agents with precise role boundaries, tool permissions, and delegation patterns.

The Claude Sub-Agent Implementation skill provides a structured framework for scaling complex AI workflows by decomposing tasks into specialized, autonomous sub-agents. It streamlines the creation and management of agent definitions under the agents/ directory, ensuring each sub-agent has a clearly defined purpose, appropriate model settings, and restricted tool access. This skill is particularly useful for high-complexity projects where multi-agent coordination is required to maintain focus, manage context efficiently, and enforce security boundaries through least-privilege tool permissions.

主な機能

01Integrated validation scripts for checking agent structure and boundary integrity

02Configurable tool access strategies and model selection guidance

03Complexity scoring framework to determine optimal agent-vs-skill architecture

040 GitHub stars

05Automated sub-agent file generation with standardized frontmatter templates

06Pre-defined templates for autonomous, read-only, and skill-integrated agents

ユースケース

01Orchestrating autonomous evaluation pipelines for testing complex AI outputs

02Implementing 'least-privilege' security by creating read-only auditing sub-agents

03Decomposing large-scale legacy code migrations into specialized domain tasks

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add yu-iskw/google-cloud-claude-code-plugin implement-sub-agents

For use in Claude.ai and ChatGPT

主な機能

01Integrated validation scripts for checking agent structure and boundary integrity

02Configurable tool access strategies and model selection guidance

03Complexity scoring framework to determine optimal agent-vs-skill architecture

040 GitHub stars

05Automated sub-agent file generation with standardized frontmatter templates

06Pre-defined templates for autonomous, read-only, and skill-integrated agents

ユースケース

01Orchestrating autonomous evaluation pipelines for testing complex AI outputs

02Implementing 'least-privilege' security by creating read-only auditing sub-agents

03Decomposing large-scale legacy code migrations into specialized domain tasks

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add yu-iskw/google-cloud-claude-code-plugin implement-sub-agents

For use in Claude.ai and ChatGPT