# My Claude Code Sub Agents BUILD THEMSELVES Source: https://www.youtube.com/watch?v=7B2HJr0Y68g Recap page: https://rapidrecap.app/video/7B2HJr0Y68g Generated: 2025-08-28T10:33:05.56+00:00 --- ## Quick Overview Claude Code's sub-agents can build other sub-agents, allowing for complex, multi-step automations and a significant scaling of engineering efforts, but require careful prompt engineering to ensure they are effective and avoid common pitfalls. **Key Points:** - Claude Code sub-agents can be used to build other sub-agents, creating complex, multi-step automations. - The "meta-agent" allows users to create new sub-agents by defining their purpose, tools, and domain. - Sub-agents operate within their own context, preventing pollution of the main conversation and keeping focus on high-level objectives. - Key benefits of sub-agents include context preservation, specialized expertise, reusability, and flexible permissions. - Proper prompt engineering is critical for effective sub-agent creation, requiring clear definitions of purpose, tools, and desired output format. - Two major mistakes to avoid are not understanding the scope of the agent and not clearly defining the expected output or prompt. - By understanding the context, model, and prompt, engineers can leverage sub-agents to scale their work efficiently. ![Screenshot at 00:00: Title card introducing the video "My Claude Code Sub Agents BUILD THEMSELVES" with a person interacting with a digital interface.](https://ss.rapidrecap.app/screens/7B2HJr0Y68g/00-00-00.png) **Context:** This video demonstrates how to build new sub-agents using Claude Code, focusing on the meta-agent's ability to generate sub-agent configurations. It explains the process of defining an agent's purpose, tools, and domain, and how to write effective prompts to ensure the agent performs its tasks correctly. The video highlights the advantages of using sub-agents for complex automations and scaling engineering efforts, while also warning about common pitfalls like context issues and improper prompt engineering. ## Detailed Analysis The video explains how to build new sub-agents using Claude Code, emphasizing the meta-agent's role in creating these agents. A sub-agent is defined by its purpose, tools, and domain, and its configuration is written in Markdown format. The meta-agent acts as an architect, taking a user's description and generating a ready-to-use sub-agent configuration. The key benefits of sub-agents are context preservation, specialized expertise, reusability, and flexible permissions, allowing engineers to tailor agents for specific tasks and scale their operations. The video stresses the importance of clear prompt engineering, including defining the agent's purpose, specifying tools, and providing a detailed system prompt. It also warns against common mistakes like not understanding the agent's scope or overloading the prompt. The process involves defining the agent's name, description, tools, and color, and then using the meta-agent to generate the configuration file. The video demonstrates how to run commands like `/all_tools` to list available tools and `/prime` to load context, and then shows how to build a new agent by providing a prompt to the meta-agent. The output format for new agents includes fields like name, description, tools, and color. The video also touches on the concept of "chaining" sub-agents for more complex tasks, creating multi-agent systems that can handle sophisticated workflows. By following these principles, engineers can effectively leverage sub-agents to improve efficiency and scale their development processes. ### Introduction - Building Sub-Agents with Claude Code: The video introduces the concept of building sub-agents with Claude Code, highlighting the meta-agent's role in generating agent configurations based on user descriptions. ### Sub-Agent Architecture and Configuration - Sub-agents are defined by their purpose, tools, and domain, with configurations written in Markdown. Key benefits include context preservation, specialized expertise, reusability, and flexible permissions. ### Prompt Engineering for Sub-Agents - Effective prompt engineering is crucial for creating functional sub-agents. This involves clearly defining the agent's purpose, specifying available tools, and providing a detailed system prompt. ### Common Mistakes and Best Practices - The video warns against common pitfalls like context issues and improper prompt engineering, emphasizing the importance of clear, concise, and specific prompts. ### Creating a New Agent - The process involves defining the agent's name, description, tools, and color, and then using the meta-agent to generate the configuration file. ### Leveraging Sub-Agents - Sub-agents can be chained together to create complex, multi-agent systems for automated workflows, allowing engineers to scale their efforts efficiently. ### Example Sub-Agents - The video briefly mentions examples like the "work-completion-summary" agent and the "tts-summary" agent to illustrate the capabilities and usage of sub-agents. ![Screenshot at 00:00: Title card introducing the video "My Claude Code Sub Agents BUILD THEMSELVES" with a person interacting with a digital interface.](https://ss.rapidrecap.app/screens/7B2HJr0Y68g/00-00-00.png) ![Screenshot at 00:06: Screenshot of a terminal window showing the command prompt with "IndyDevDan: claude-code-hooks-mastery %" and text "you fire up Claude Code,".](https://ss.rapidrecap.app/screens/7B2HJr0Y68g/00-00-06.png) ![Screenshot at 00:07: Terminal window displaying commands and descriptions for various agents like "agentic\_starter\_kit", "ai\_docs", and "all\_tools".](https://ss.rapidrecap.app/screens/7B2HJr0Y68g/00-00-07.png) ![Screenshot at 00:10: Terminal window showing the command "\> /cook" and its description "Run these 8 sub agent tasks simultaneously in parallel: \(project\)".](https://ss.rapidrecap.app/screens/7B2HJr0Y68g/00-00-10.png) ![Screenshot at 00:15: Terminal window showing "/cook is running..." and a list of sub-agents being initialized.](https://ss.rapidrecap.app/screens/7B2HJr0Y68g/00-00-15.png) ![Screenshot at 00:21: Terminal window listing "Update Todos" with bullet points for "crypto-coin-analyzer", "crypto-market-agent", and "codebase-consultant-cc-hooks".](https://ss.rapidrecap.app/screens/7B2HJr0Y68g/00-00-21.png) ![Screenshot at 00:33: Terminal window displaying "Code is a commodity." and "Your fine tune prompts can be VALUABLE.".](https://ss.rapidrecap.app/screens/7B2HJr0Y68g/00-00-33.png) ![Screenshot at 00:38: Diagram illustrating the "Claude Code Sub Agent Flow" with "User Prompt" and "Primary Agent" nodes.](https://ss.rapidrecap.app/screens/7B2HJr0Y68g/00-00-38.png) ![Screenshot at 01:08: Diagram showing the "Claude Code Sub Agent Flow" with "User Prompt", "Primary Agent", "debugger", "data-scientist", and "optimizer" nodes.](https://ss.rapidrecap.app/screens/7B2HJr0Y68g/00-01-08.png) ![Screenshot at 01:23: Diagram showing the "Claude Code Sub Agent Flow" with "User Prompt", "Primary Agent", "debugger", "data-scientist", "optimizer", and "code-reviewer" nodes connected in a chain-like structure.](https://ss.rapidrecap.app/screens/7B2HJr0Y68g/00-01-23.png)