skill-developer

Claude Code skills use YAML frontmatter in SKILL.md to define metadata, trigger patterns via UserPromptSubmit and PreToolUse hooks, and skill-rules.json for enforcement configuration. This automated approach ensures reliable activation and guardrail application across skill files.

UserPromptSubmit hooks trigger suggestions early in user queries, while PreToolUse hooks enforce guardrails before tool execution. Use UserPromptSubmit for contextual guidance and PreToolUse for safety-critical domain constraints and session-aware rule enforcement.

skill-rules.json defines rules with enforcement levels controlling how strictly triggers apply. Structure includes trigger conditions, action types, and session tracking to avoid redundant notifications. Progressive disclosure guidelines prevent overwhelming users while maintaining guardrail integrity.

Yes. Create a skill folder with SKILL.md frontmatter, add optional resources, populate skill-rules.json, and run tests to verify trigger patterns activate correctly across your skill files before live use.

The 500-line rule establishes a complexity ceiling for skill maintainability. Skill design encompasses frontmatter guidelines, hook mechanics, progressive disclosure, and guardrail types to ensure scalable, reliable auto-activation without exceeding functional limits.

Session tracking in skill-rules.json tracks which guardrails fired within a single session. This prevents repeated enforcement of the same rule, improving user experience while maintaining safety constraints across Claude Code interactions.