skill-developer

Structure a Claude Skill by defining YAML frontmatter with metadata, following the 500-line rule to keep skill logic concise, and using progressive disclosure to layer complexity for AI agent processing. This ensures your skill triggers correctly and integrates seamlessly into auto-injection systems.

Trigger patterns are keyword-based and intent-driven rules that determine when your skill auto-activates. Define them by analyzing user intent and setting specific keywords in your skill configuration, enabling Claude Code to invoke your skill without manual context loading.

Configure skill-rules.json by specifying skill type (domain or guardrail), defining trigger patterns, setting hook mechanisms like UserPromptSubmit, and establishing naming conventions and scope rules. This configuration file controls how your skill behaves within the Claude Code environment.

Test and debug by validating YAML frontmatter syntax, verifying trigger patterns activate correctly, checking hook injection points, and running integration tests against the self-healing guidelines. This ensures your skill functions as expected when auto-injected.

Yes, session tracking and documentation structure are part of Anthropic best practices for Claude Skills. Implement them in your skill-rules.json and SKILL.md to enable proper state management, maintain clarity for AI agents, and ensure consistent skill behavior across sessions.

Yes, the UserPromptSubmit hook enables seamless skill injection by triggering your skill logic when a user submits a prompt. Configure it in skill-rules.json to intercept and process user input according to your skill's domain logic and guardrail rules.