Can a subagent spawn another subagent?

No, subagents are restricted from spawning other subagents to prevent infinite nesting and ensure predictable performance.

Where are custom subagents stored?

Custom agents are stored as .md files in .claude/agents/ for specific projects or in ~/.claude/agents/ for global access across all your projects.

What are subagent hooks?

Hooks are lifecycle-scoped actions that run before or after a tool is used, allowing you to log activity or block specific commands for safety.

What is a Claude Code subagent?

A subagent is a specialized AI assistant that Claude Code delegates tasks to. It operates in its own context window with specific tools and instructions to handle domain-specific work.

Can I restrict what tools a subagent can use?

Yes, you can define exactly which tools an agent can access using the 'tools' or 'disallowedTools' fields in the agent's Markdown frontmatter.

Subagent Architect

Name: Subagent Architect
Author: hgeldenhuys

byhgeldenhuys

•

开发者工具

Enables the creation and management of specialized autonomous subagents within Claude Code for domain-specific tasks and automated workflows.

关于

The Subagent Architect skill provides a comprehensive framework for building custom AI assistants tailored to specific project needs. By defining specialized agents with unique system prompts, restricted tool access, and specific model configurations, developers can delegate complex tasks like security auditing, database management, or test automation without cluttering the main conversation context. This skill covers everything from basic YAML frontmatter configuration to advanced lifecycle hooks for tool validation and proactive delegation patterns.

主要功能

Multi-model support allowing agents to use Opus, Sonnet, or Haiku
Interactive /agents management interface for creation and editing
Custom tool restriction through allowlists and denylists
1 GitHub stars
Lifecycle hooks for pre/post-tool execution and command validation
Project-level and global user-level agent scoping

使用场景

Creating an automated test-runner that proactively fixes code failures during development
Building a read-only code reviewer to scan pull requests for security vulnerabilities
Setting up restricted database agents that can only execute safe SELECT queries

关于

主要功能

Multi-model support allowing agents to use Opus, Sonnet, or Haiku
Interactive /agents management interface for creation and editing
Custom tool restriction through allowlists and denylists
1 GitHub stars
Lifecycle hooks for pre/post-tool execution and command validation
Project-level and global user-level agent scoping

使用场景

Creating an automated test-runner that proactively fixes code failures during development
Building a read-only code reviewer to scan pull requests for security vulnerabilities
Setting up restricted database agents that can only execute safe SELECT queries