What are the naming requirements for agents?

Agent identifiers must be between 3-50 characters, use lowercase letters, numbers, and hyphens only, and must start and end with an alphanumeric character.

Can I limit which tools an agent can use?

Yes, the 'tools' field in the agent's frontmatter allows you to restrict the agent to a specific set of tools, adhering to the principle of least privilege for safer execution.

How do I trigger a specific agent?

Agents are triggered via the 'description' field in their frontmatter. This field should include specific 'Use this agent when' conditions and multiple concrete examples of user requests.

What is the recommended model for an agent?

Using 'inherit' is generally recommended so the agent uses the same model as the parent session, though you can specify sonnet, opus, or haiku if needed.

What is a Claude Code agent?

An agent is an autonomous subprocess within a Claude Code plugin designed to handle complex, multi-step tasks independently using specific tools and specialized system prompts.

Agent Development Guide

Name: Agent Development Guide
Author: liminal-ai

byliminal-ai

0•

Learning & Documentation

Provides comprehensive guidance and best practices for building autonomous agents within Claude Code plugins.

This skill serves as an essential manual for developers looking to extend Claude's capabilities through custom agents. It provides standardized templates and best practices for every stage of development, from YAML frontmatter configuration and triggering logic to system prompt engineering and tool restriction strategies. By following these patterns, developers can create reliable, autonomous subprocesses capable of handling complex, multi-step workflows with precision and visual clarity in the Claude Code interface.

Key Features

01Validation rules for agent identifiers, descriptions, and prompt lengths

02Expert system prompt design patterns for autonomous personas

03Standardized agent file structure and YAML frontmatter templates

04Best practices for triggering logic using detailed description examples

050 GitHub stars

06Guidelines for tool permission sets and model selection optimization

Use Cases

01Building specialized subagents for code reviews, security analysis, or test generation

02Optimizing agent triggering conditions to ensure Claude invokes the correct sub-process proactively

03Designing robust system prompts that enforce specific methodologies and output formats

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add liminal-ai/code-steward agent-development

For use in Claude.ai and ChatGPT

Key Features

01Validation rules for agent identifiers, descriptions, and prompt lengths

02Expert system prompt design patterns for autonomous personas

03Standardized agent file structure and YAML frontmatter templates

04Best practices for triggering logic using detailed description examples

050 GitHub stars

06Guidelines for tool permission sets and model selection optimization

Use Cases

01Building specialized subagents for code reviews, security analysis, or test generation

02Optimizing agent triggering conditions to ensure Claude invokes the correct sub-process proactively

03Designing robust system prompts that enforce specific methodologies and output formats

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add liminal-ai/code-steward agent-development

For use in Claude.ai and ChatGPT