creating-claude-agents

Creating a Claude agent requires defining frontmatter with metadata like name, description, model, and allowed tools, then validating the markdown file against the required JSON schema. This Skill automates that validation and provides templates ensuring your agent meets all structural and technical requirements for .claude/agents/ files.

Best practices include defining clear persona instructions, selecting the appropriate Claude model, specifying which tools the agent can access, and structuring the agent file with a required H1 heading and consistent formatting. This Skill provides templates and guidance for each component to ensure agents are optimized and consistent.

Yes. This Skill validates existing agent markdown files in .claude/agents/ against the required schema, checking name patterns, description length, model values, tool definitions, and overall JSON conformance. It identifies inconsistencies and guides you to fix structural or technical issues.

If frontmatter violates schema rules—such as invalid name patterns, missing descriptions, unsupported model values, or malformed tool access lists—this Skill flags those errors and explains what needs correction to bring your agent into compliance.

This Skill guides you through agent creation for specialized use cases by ensuring you define the correct persona, select security-appropriate models and tools, structure the markdown file properly, and validate against the schema. Follow the quickstart template to build a focused, production-ready agent.

This Skill includes ready-to-use templates for common agent types that include properly formatted frontmatter, persona definitions, and process workflows. You customize the template for your use case rather than building frontmatter from scratch.