infra:make-subagent

Creating specialized Claude Code sub-agents requires configuring a specific file structure, frontmatter, and system prompts. This process involves defining the agent's role, selecting appropriate tools like grep or bash, and formatting system prompts using XML tags to ensure proper execution.

Selecting AI models for custom Claude Code agents depends on task complexity, choosing between sonnet, opus, and haiku. You must also configure appropriate tools like grep and bash based on the specific data processing or code analysis requirements defined in the agent's workflow.

The correct file structure for Claude Code sub-agents requires specific frontmatter configurations and XML tags for system prompts. Adhering to these formatting constraints ensures proper agent initialization, tool selection, and background execution without requiring direct user interaction.

Claude Code sub-agents support background execution and parallel agent patterns. Designing the workflow involves differentiating between main chat responsibilities and agent tasks, allowing multiple specialized agents to process code analysis or data simultaneously.

Limitations when designing Claude Code sub-agents include strict constraints on user interaction and mandatory use of XML tags for system prompts. Agents must adhere to these structural rules and cannot operate outside their configured tool selections and defined roles.