Skill Developer Guide

Purpose

Comprehensive guide for creating and managing skills in Claude Code, following the Agent Skills Specification and Anthropic's best practices including the 500-line rule and progressive disclosure pattern.

Agent Skills Specification (agentskills.io)

Skills in this project follow the open Agent Skills Specification. Key requirements:

Directory Structure

skills/{skill-name}/
├── SKILL.md           # Required - main skill file
├── references/        # Optional - detailed documentation
├── scripts/           # Optional - executable code (Python, Bash, JS)
└── assets/            # Optional - templates, images, data files

SKILL.md Frontmatter Schema

Required Fields:

Optional Fields:

Progressive Disclosure Model

1. Metadata (~100 tokens) - loads at startup
2. Full instructions (<5000 tokens recommended) - loads on activation
3. Supporting files - loads as needed

Validation

Use skills-ref tool or validate manually:

• File must be named SKILL.md (uppercase)
• Name in frontmatter must match parent directory
• Description should include trigger keywords

When to Use This Skill

Automatically activates when you mention:

• Creating or adding skills
• Modifying skill triggers or rules
• Understanding how skill activation works
• Debugging skill activation issues
• Claude Code best practices
• Progressive disclosure
• YAML frontmatter
• 500-line rule

System Overview

Skill Activation Approaches

There are several approaches to skill activation, from simple to advanced:

1. Description-Based Activation (Default)

• Claude reads skill descriptions from frontmatter at session start
• Activates skills when user prompts match description keywords
• No additional configuration needed
• Works out of the box with the agentskills.io spec

2. Example Implementation: Hook-Based Activation

For more sophisticated activation, you can build a hook system. This is an advanced pattern:

UserPromptSubmit Hook (Proactive Suggestions)

• Trigger: BEFORE Claude sees user's prompt
• Purpose: Suggest relevant skills based on keywords + intent patterns
• Method: Injects formatted reminder as context (stdout to Claude's input)

Stop Hook (Gentle Reminders)

• Trigger: AFTER Claude finishes responding
• Purpose: Gentle reminder to self-assess patterns in code written
• Method: Analyzes edited files for patterns, displays reminder if needed

Configuration for Hook Systems:

A rules file (e.g., skill-rules.json) can define:

• All skills and their trigger conditions
• Enforcement levels (block, suggest, warn)
• File path patterns (glob)
• Content detection patterns (regex)
• Skip conditions (session tracking, file markers, env vars)

Skill Types

1. Guardrail Skills

Purpose: Enforce critical best practices that prevent errors

Characteristics:

• Type: "guardrail"
• Enforcement: "block"
• Priority: "critical" or "high"
• Block file edits until skill used
• Prevent common mistakes
• Session-aware (don't repeat nag in same session)

Examples:

• database-verification - Verify table/column names before queries
• frontend-dev-guidelines - Enforce React/TypeScript patterns

When to Use:

• Mistakes that cause runtime errors
• Data integrity concerns
• Critical compatibility issues

2. Domain Skills

Purpose: Provide comprehensive guidance for specific areas

Characteristics:

• Type: "domain"
• Enforcement: "suggest"
• Priority: "high" or "medium"
• Advisory, not mandatory
• Topic or domain-specific
• Comprehensive documentation

Examples:

• backend-dev-guidelines - Node.js/Express/TypeScript patterns
• frontend-dev-guidelines - React/TypeScript best practices
• drupal-search-api - Search API configuration guidance

When to Use:

• Complex systems requiring deep knowledge
• Best practices documentation
• Architectural patterns
• How-to guides

Quick Start: Creating a New Skill

Step 1: Create Skill File

Location: skills/{skill-name}/SKILL.md

Template:

---
name: my-new-skill
description: Brief description including keywords that trigger this skill. Mention topics, file types, and use cases. Be explicit about trigger terms.
---

# My New Skill

## Purpose
What this skill helps with

## When to Use
Specific scenarios and conditions

## Key Information
The actual guidance, documentation, patterns, examples

Best Practices (per agentskills.io spec):

• Name: 1-64 chars, lowercase alphanumeric + hyphens, must match directory name
• Description: 1-1024 chars, include ALL trigger keywords/phrases
• Content: Under 500 lines - use references/ directory for details
• Examples: Real code examples
• Structure: Clear headings, lists, code blocks

Step 2: (Optional) Add Trigger Rules

If using a hook-based activation system, add rules:

Basic Template:

{
  "my-new-skill": {
    "type": "domain",
    "enforcement": "suggest",
    "priority": "medium",
    "promptTriggers": {
      "keywords": ["keyword1", "keyword2"],
      "intentPatterns": ["(create|add).*?something"]
    }
  }
}

Step 3: Test Activation

Verify your skill activates correctly:

• Try prompts that should trigger it
• Try prompts that should NOT trigger it
• Refine description keywords based on results

Step 4: Follow Anthropic Best Practices

• Keep SKILL.md under 500 lines
• Use progressive disclosure with reference files
• Add table of contents to reference files > 100 lines
• Write detailed description with trigger keywords
• Test with 3+ real scenarios before documenting
• Iterate based on actual usage

Enforcement Levels

BLOCK (Critical Guardrails)

• Physically prevents Edit/Write tool execution
• Claude sees message and must use skill to proceed
• Use For: Critical mistakes, data integrity, security issues

Example: Database column name verification

SUGGEST (Recommended)

• Reminder injected before Claude sees prompt
• Claude is aware of relevant skills
• Not enforced, just advisory
• Use For: Domain guidance, best practices, how-to guides

Example: Frontend development guidelines

WARN (Optional)

• Low priority suggestions
• Advisory only, minimal enforcement
• Use For: Nice-to-have suggestions, informational reminders

Rarely used - most skills are either BLOCK or SUGGEST.

Skip Conditions & User Control

1. Session Tracking

Purpose: Don't nag repeatedly in same session

How it works:

• First edit: Hook blocks, updates session state
• Second edit (same session): Hook allows
• Different session: Blocks again

2. File Markers

Purpose: Permanent skip for verified files

Marker: // @skip-validation

Usage:

// @skip-validation
import { PrismaService } from './prisma';
// This file has been manually verified

NOTE: Use sparingly - defeats the purpose if overused

3. Environment Variables

Purpose: Emergency disable, temporary override

Global disable:

export SKIP_SKILL_GUARDRAILS=true  # Disables ALL blocks

Skill-specific:

export SKIP_DB_VERIFICATION=true

Testing Checklist

When creating a new skill, verify:

• Skill file created in skills/{name}/SKILL.md
• Proper frontmatter with name and description
• Keywords tested with real prompts
• No false positives in testing
• No false negatives in testing
• SKILL.md under 500 lines
• Reference files created if needed
• Table of contents added to files > 100 lines

Advanced Topics

For teams building sophisticated activation systems, consider:

• Trigger types: Keywords, intent patterns, file path globs, content regex
• Hook mechanisms: UserPromptSubmit, PreToolUse, Stop event hooks
• Session state management: Tracking which skills have been acknowledged
• Performance: Keep hook execution under 200ms
• Pattern libraries: Reusable regex and glob patterns

These topics are documented in optional reference files:

• TRIGGER_TYPES.md - Complete trigger type guide
• SKILL_RULES_REFERENCE.md - Rules schema documentation
• HOOK_MECHANISMS.md - Hook internals deep dive
• TROUBLESHOOTING.md - Debugging activation issues
• PATTERNS_LIBRARY.md - Ready-to-use pattern collection

Quick Reference Summary

Create New Skill (4 Steps)

1. Create skills/{name}/SKILL.md with frontmatter
2. Test activation with real prompts
3. Refine description keywords based on testing
4. Keep SKILL.md under 500 lines

Trigger Types

• Keywords: Explicit topic mentions (in description)
• Intent: Implicit action detection (advanced, via hooks)
• File Paths: Location-based activation (advanced, via hooks)
• Content: Technology-specific detection (advanced, via hooks)

Enforcement

• BLOCK: Critical only, prevents tool execution
• SUGGEST: Most common, advisory context injection
• WARN: Advisory, rarely used

Skip Conditions

• Session tracking: Automatic (prevents repeated nags)
• File markers: // @skip-validation (permanent skip)
• Env vars: SKIP_SKILL_GUARDRAILS (emergency disable)

Anthropic Best Practices & agentskills.io Spec

• 500-line rule: Keep SKILL.md under 500 lines
• Progressive disclosure: Use references/ directory for detailed docs
• Table of contents: Add to reference files > 100 lines
• One level deep: Don't nest references deeply
• Rich descriptions: Include all trigger keywords (max 1024 chars per spec)
• Name = Directory: Skill name in frontmatter must match parent directory
• Test first: Build 3+ evaluations before extensive documentation
• Spec URL: https://agentskills.io/specification

Troubleshoot

Test skill activation by trying prompts that should and should not trigger the skill. Refine description keywords iteratively.

Reference Files

Deep-dive documentation for the hook-based activation pattern described above (the advanced, optional architecture — description-based activation needs none of this):

• references/TRIGGER_TYPES.md - Keyword, intent, file-path, and content trigger types in detail
• references/SKILL_RULES_REFERENCE.md - Complete skill-rules.json schema
• references/HOOK_MECHANISMS.md - How UserPromptSubmit/Stop hooks work
• references/TROUBLESHOOTING.md - Debugging skill activation
• references/PATTERNS_LIBRARY.md - Reusable trigger/rule patterns
• references/ADVANCED.md - Advanced topics

Related Files

Configuration:

• skills/*/SKILL.md - Skill content files

Specification:

• https://agentskills.io/specification - Agent Skills Spec

Attribution

The hook-based activation architecture and the reference files above are adapted from diet103/claude-code-infrastructure-showcase (MIT License).