subagent-creator

Create a sub-agent by writing a Markdown file with YAML frontmatter at .claude/agents/<name>.md or ~/.claude/agents/<name>.md. Define required fields (name, description) and optional fields (tools, model, permissionMode, skills), then add your system prompt below the frontmatter. Use assets/subagent-template.md as a starting point.

System prompts define the sub-agent's role, responsibilities, behavior, and output format. You can configure tool access, model selection, permission modes, and skill auto-loading to tailor the agent for domain-specific tasks like code review, debugging, or data analysis.

Yes. Each sub-agent's YAML frontmatter includes an optional tools field where you specify which tools that agent can access. This lets you create specialized agents—one for code review with linting tools, another for data work with analysis tools—all in the same project.

Sub-agents support both scopes. Store them in .claude/agents/ for project-level scope (higher priority) or ~/.claude/agents/ for user-level scope (lower priority). Project-scoped agents override user-scoped ones with the same name.

Sub-agents are specialized, focused instances with custom system prompts and restricted tool access for automating domain-specific tasks. The main workflow is general-purpose; sub-agents let you isolate behavior, permissions, and responsibility boundaries for automation and multi-step processes.

Yes. Sub-agents support an optional skills field in their YAML frontmatter that auto-loads reusable templates and skill configurations. This lets you compose specialized agents from existing Skill definitions without rewriting logic.