What is a Claude Code sub-agent?

A sub-agent is a specialized instance of Claude with a specific persona, toolset, and instructions designed to handle a niche task within your development workflow.

Can I restrict the tools available to a sub-agent?

Yes, the Sub-Agent Creator allows you to define read-only, code-ops, or custom tool access patterns to ensure security and prevent unauthorized modifications.

Does this skill help with prompt engineering?

Yes, it includes a dedicated step for token optimization and provides a structured system prompt template to ensure high-quality outputs while minimizing resource usage.

Where are sub-agents stored?

Sub-agents can be stored at the project level in .claude/agents/ for team sharing or at the user level in ~/.claude/agents/ for personal use across multiple projects.

How does the proactive trigger work?

By defining specific trigger phrases and conditions in the agent's description, Claude can automatically delegate tasks to the sub-agent when relevant patterns are detected in the conversation.

Sub-Agent Creator for Claude Code

Name: Sub-Agent Creator for Claude Code
Author: cskiro

bycskiro

•

Productivity & Workflow

Automates the generation of specialized, proactive Claude Code sub-agents following official Anthropic architectural patterns.

Sub-Agent Creator is a specialized capability designed to streamline the development of task-specific sub-agents within the Claude Code environment. By guiding users through a structured 6-step architectural workflow—covering intent extraction, persona design, and token optimization—the skill ensures every sub-agent is correctly configured with required YAML frontmatter and appropriate tool restrictions. It is particularly effective for teams needing consistent, domain-specific assistants such as security reviewers or DevOps experts, enabling efficient task delegation and proactive behavior through automated trigger patterns.

Key Features

01Automated YAML frontmatter and tool configuration generation

02Structured 6-step architect workflow for consistent agent quality

03Proactive auto-triggering behavior implementation

04Token-efficient system prompt optimization techniques

05Dual-example training patterns for precise agent invocation

062 GitHub stars

Use Cases

01Creating team-shared project-level agents for specialized workflows

02Setting up security-sensitive agents with restricted tool access

03Generating domain-specific agents like code reviewers or debuggers

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add cskiro/claudex sub-agent-creator

For use in Claude.ai and ChatGPT

Download Skill