skill-developer

Auto-activation rules for Claude skills use trigger patterns, hook mechanisms, and enforcement levels (block, suggest, warn) configured in YAML frontmatter and JSON skill-rules. Define fileTriggers for file-path matching, promptTriggers for intent detection, and skipConditions to refine when rules apply, then test across guardrail and domain skill types.

Hook mechanisms like UserPromptSubmit and PreToolUse intercept Claude Code workflow events to enforce rules at the right moment. Implement them by specifying hook names in your skill configuration, mapping them to trigger patterns, and defining the enforcement action (block, suggest, or warn) that executes when the hook fires.

Trigger patterns detect when a skill should activate by matching keywords, file types, or prompt content. Configure them using promptTriggers for keyword and intent matching, fileTriggers for file paths and extensions, and skipConditions to exclude specific scenarios—all defined in YAML frontmatter alongside enforcement levels.

Yes, session tracking is supported in skill-developer for monitoring state and trigger history across interactions. Configure it in your skill-rules JSON to track enforcement decisions, trigger occurrences, and user context, enabling stateful governance and multi-interaction rule refinement.

Guardrail skills enforce constraints and safety policies with block or suggest actions across all interactions. Domain skills provide specialized logic for specific workflows. Skill-developer supports both types with distinct configuration patterns, enforcement levels, and hook integration appropriate to each skill's governance scope.

Test trigger patterns by creating test scaffolds that simulate UserPromptSubmit and PreToolUse hook events, validate that fileTriggers and promptTriggers match your intended scenarios, and verify enforcement actions execute correctly. Skill-developer includes testing support to catch configuration errors and edge cases early.