skill-developer

Create a new skill by adding a SKILL.md file with YAML frontmatter to .claude/skills/{name}/, then register it in .claude/skills/skill-rules.json. Define trigger patterns, hooks, and guardrail enforcement rules in the frontmatter to activate the skill based on keywords, file paths, or content patterns.

SKILL.md frontmatter specifies trigger types (keywords, intent patterns, file paths, content patterns), activation rules, guardrail enforcement levels, hook mechanisms, and session tracking config. The structure guides progressive disclosure via reference files to avoid overwhelming documentation while maintaining complete skill definitions.

Use provided CLI scripts to validate trigger matching against keywords, intent patterns, file paths, and content patterns. Test guardrail enforcement and hook execution across domain and guardrail skills to ensure consistent activation behavior before deploying skills to the Claude Code environment.

Yes. Progressive disclosure embeds reference files in SKILL.md to reveal deeper details on demand, avoiding cognitive overload. Link to versioning, hot-reload concepts, analytics hooks, and automated testing scaffolds separately so users access only what they need at each stage.

Keyword triggers activate on exact or fuzzy term matches within user input; intent pattern triggers recognize broader semantic intent and workflow context. Combine both with file path and content pattern triggers to build layered, context-aware skill activation across guardrail and domain workflows.

Versioning and hot-reload concepts are satisfied through skill-rules.json configuration and SKILL.md structure. The framework supports iterative skill updates and testing scaffolds to validate changes without disrupting active sessions, enabling consistent skill evolution across the Claude Code environment.