Where are the created agents stored?

Agents are stored as Markdown files in ~/.claude/agents/ for global (user-level) access or within a project's .claude/agents/ directory for repo-specific behavior.

Does this skill follow Opus 4.5 prompting best practices?

Yes, it includes specific guidelines for Opus 4.5, such as using imperative language, avoiding aggressive keywords, and including anti-over-engineering guidance.

What is the difference between an Agent and a Skill in Claude Code?

An Agent changes who Claude is (persona, tools, model), while a Skill changes what Claude knows (knowledge, procedures, scripts). Agents can load skills, but skills cannot become agents.

Can I restrict which tools an agent can use?

Yes, the Agent Creator allows you to define a whitelist of tools in the YAML frontmatter, preventing the agent from using unauthorized capabilities like Bash or Write.

Why should I use the Haiku model for some agents?

Haiku is recommended for simple, repetitive tasks like quick searching or basic data extraction to optimize for speed and significantly reduce token costs.

Agent Creator

Name: Agent Creator
Author: GGPrompts

byGGPrompts

•

生産性とワークフロー

Streamlines the creation and configuration of specialized Claude Code agents with custom personas, tool restrictions, and model selections.

概要

The Agent Creator skill provides a comprehensive framework for building and optimizing Claude Code agents, which are specialized personas that fundamentally change Claude's behavior for specific tasks. It guides users through the entire lifecycle of agent development—from defining tool whitelists and model selection (Haiku, Sonnet, or Opus) to crafting high-performance system prompts using Opus 4.5 best practices. This skill is essential for developers looking to automate repetitive workflows, restrict tool usage for security, or create domain-specific orchestrators that can delegate tasks to sub-agents.

主な機能

Automated agent file initialization with pre-configured scaffolding
2 GitHub stars
Opus 4.5 compliant system prompt engineering guidelines
Support for both project-level and global user-level agent configurations
Guided tool restriction and whitelist management for enhanced security
Model optimization recommendations for cost and performance balancing

ユースケース

Building a specialized Code Reviewer agent with restricted tool access
Creating a lightweight Haiku-based agent for fast repository searching and indexing
Developing an Orchestrator agent to manage complex sub-tasks and skill integrations

概要

主な機能

Automated agent file initialization with pre-configured scaffolding
2 GitHub stars
Opus 4.5 compliant system prompt engineering guidelines
Support for both project-level and global user-level agent configurations
Guided tool restriction and whitelist management for enhanced security
Model optimization recommendations for cost and performance balancing

ユースケース

Building a specialized Code Reviewer agent with restricted tool access
Creating a lightweight Haiku-based agent for fast repository searching and indexing
Developing an Orchestrator agent to manage complex sub-tasks and skill integrations