subagent-creator

You create Claude Code subagents by placing a markdown file named <agent-name>.md in the desired scope directory, adding YAML frontmatter with required fields, and writing a focused system prompt in the body to configure delegation.

Configuring tools and permissions for Claude Code subagents is handled by defining tool whitelisting, models, and permission modes in the YAML frontmatter. This standardizes agent capabilities and enforces safe execution boundaries across projects.

You validate subagent configuration by checking naming conventions, scope uniqueness, and alignment with the intended trigger. This verification ensures the subagent behaves correctly before you restart the session to load the file.

Yes, hook-driven validation is supported during subagent creation. You can define hooks in the YAML frontmatter to enforce standardized configuration rules and ensure safe execution before the agent is loaded into the environment.

Setting up task delegation requires defining scope selection, models, and tools in the YAML frontmatter. This determines what the subagent can access and how it processes delegated tasks within your projects.

Claude Code subagents are loaded on session restart. If your subagent is not loading, verify the YAML frontmatter fields, ensure the file is named <agent-name>.md in the correct scope directory, and restart the session to load the new configuration.