agent-builder

To build Claude Code subagents, you create a file under .claude/agents/ with proper frontmatter, a clear system prompt, and minimal tooling configuration. This skill provides step-by-step templates for scaffolding agents with precise delegation patterns.

Configuring tool access for subagents involves granting minimal permissions like Read, Write, Bash, or Grep based on the task. This skill provides guidance on tool access patterns to ensure controlled tooling and maintainable agent behavior.

Writing system prompts for subagents requires defining a clear system prompt with documented output formats and precise frontmatter. This skill offers templates to structure prompts that enforce delegation and maintainability across software engineering tasks.

Yes, you can create specialized subagents for code review, debugging, and data science tasks. This skill supports designing and refining agents for specific use cases by applying best practices for delegation and chaining.

Model selection is a core part of subagent creation that affects task performance. This skill provides guidance on choosing appropriate models alongside configuring frontmatter and system prompts for optimal delegation patterns.

Delegation patterns fail when subagents lack precise frontmatter, clear system prompts, or minimal tool access. This skill helps debug and refine existing subagents by enforcing structured prompts and controlled tooling configurations.