What is the difference between BLOCK and SUGGEST enforcement?

BLOCK enforcement physically prevents a tool from executing until the user acknowledges the skill, ideal for critical safety. SUGGEST enforcement injects context into the prompt as a helpful reminder without interrupting the workflow.

What is the 500-line rule for Claude Code skills?

The 500-line rule is a best practice recommending that SKILL.md files stay under 500 lines. This ensures the AI maintains focus and doesn't exhaust its context window with unnecessary information, utilizing progressive disclosure for deeper details.

How does the UserPromptSubmit hook work?

The UserPromptSubmit hook acts proactively. It analyzes the user's prompt for keywords or intent patterns before Claude sees it, then injects relevant skill documentation as context to guide Claude's response.

Can I skip skill validation for specific files?

Yes, you can use the '// @skip-validation' marker within a file to permanently bypass skill guardrails, or use environment variables like SKIP_SKILL_GUARDRAILS for emergency global disables.

Claude Code Skill Developer

Name: Claude Code Skill Developer
Author: JNLei

byJNLei

•

生産性とワークフロー

Streamlines the creation and management of custom Claude Code skills using Anthropic best practices and hook-based architectures.

概要

The Skill Developer is a specialized framework designed for engineers looking to extend Claude Code's capabilities. It provides structured guidance on implementing the 'Two-Hook Architecture,' allowing for proactive suggestions via UserPromptSubmit and critical guardrails through PreToolUse. By enforcing standards like the 500-line rule and progressive disclosure, it ensures that custom skills remain performant and contextually relevant. Whether you are building high-priority database safety checks or advisory frontend style guides, this skill offers the implementation patterns, trigger logic, and debugging tools necessary to build a robust internal developer platform within Claude.

主な機能

Comprehensive trigger configuration (Keywords, Intent, File Paths, Content Patterns)
Automated testing workflows for skill activation and verification
Session tracking and skip conditions to prevent tool fatigue
6 GitHub stars
Enforcement level management including BLOCK, SUGGEST, and WARN
Two-Hook Architecture implementation for proactive and reactive guidance

ユースケース

Creating critical guardrails to prevent destructive database or infrastructure changes
Developing specialized AI workflows for proprietary internal APIs and frameworks
Establishing organizational coding standards through custom domain skills

概要

主な機能

Comprehensive trigger configuration (Keywords, Intent, File Paths, Content Patterns)
Automated testing workflows for skill activation and verification
Session tracking and skip conditions to prevent tool fatigue
6 GitHub stars
Enforcement level management including BLOCK, SUGGEST, and WARN
Two-Hook Architecture implementation for proactive and reactive guidance

ユースケース

Creating critical guardrails to prevent destructive database or infrastructure changes
Developing specialized AI workflows for proprietary internal APIs and frameworks
Establishing organizational coding standards through custom domain skills