skill-developer

Claude Code skills use trigger patterns and hooks to auto-activate automation. Define triggers in `skill-rules.json` to match keywords or file patterns, configure enforcement levels (block, suggest, warn), and bind logic to `UserPromptSubmit` or `PreToolUse` hooks. Test activation by simulating prompts or file edits that match your trigger patterns.

Structure guardrail skills by defining `fileTriggers` and `contentPatterns` targeting specific file types (e.g., SQL files), setting enforcement to `block` or `suggest`, and configuring session-aware checks via `skill-rules.json`. Include references to document trigger syntax and progressive disclosure rules to keep skills under the 500-line limit.

Trigger patterns match keywords, file paths, and content within prompts or code context. Intent patterns detect semantic purpose. Hooks like `UserPromptSubmit` intercept prompts for evaluation; `PreToolUse` validates tool requests. Session tracking enables stateful enforcement across multiple interactions within a single session.

Debug skills by testing trigger matching with sample prompts and files, inspecting `skill-rules.json` configuration for enforcement levels, and validating hook orchestration across `UserPromptSubmit` and `PreToolUse` stages. Use `jq` to parse rule configurations and trace progressive disclosure outputs.

Skills must stay under 500 lines, follow progressive disclosure patterns to avoid token bloat, and use session-aware guardrails for consistency. Trigger matching must be precise to avoid false positives; enforcement levels must align with use case severity to balance automation with user control.

Yes. `npx` manages skill dependencies and tooling, `tsx` executes TypeScript-based skill logic, and `jq` parses `skill-rules.json` configurations during development and testing. These dependencies enable rapid iteration on trigger patterns and hook logic.