Claude Custom Subagents (Build Your Own AI Workers) | CampusX
Machine-readable: Markdown · JSON API · Site index
Описание видео
This video dives into Custom Subagents in Claude Code and shows how to go beyond built-in agents by creating specialized, task-specific workflows.
It starts by explaining why custom subagents are needed, especially when built-in agents cannot follow project-specific rules (like security audits or testing standards). Then it breaks down how subagents are created using markdown files, YAML configuration, tools, and system prompts.
The video also covers how to trigger subagents (automatically and manually using slash commands) and integrates everything into a real project. A full workflow is built using multiple subagents for testing, code review, and validation—demonstrating how to create scalable, modular AI development pipelines.
Chapters
00:00:00 - Introduction & Recap of Subagents
00:01:14 - Why Custom Subagents are Needed (vs Built-in)
00:02:26 - Real Example: Security Audit Use Case
00:04:17 - Concept: Specialized vs Generic Agents
00:05:47 - How to Create Custom Subagents (Core Idea)
00:06:30 - Subagent Structure (Markdown + YAML + Instructions)
00:08:10 - Project vs Personal Subagents
00:09:00 - How Subagents are Triggered (Auto vs Manual)
00:10:00 - Using Slash Commands to Trigger Subagents
00:10:48 - Project Setup & Feature Goal (Profile Page Enhancement)
00:12:00 - Spec-Driven Workflow Recap
00:14:00 - New Workflow: Testing + Code Review Added
00:15:00 - Designing Subagent Workflow (Testing Pipeline)
00:15:30 - Test Writer + Test Runner Subagents
00:16:30 - Code Review Subagents (Security + Quality)
00:18:00 - Practical: Creating Custom Subagents
00:19:00 - Creating Test Writer Subagent
00:26:00 - Creating Test Runner Subagent
00:28:50 - Connecting Subagents via Slash Command