Skip to content

Latest commit

 

History

History
426 lines (316 loc) · 9.78 KB

File metadata and controls

426 lines (316 loc) · 9.78 KB

🎨 Skills Commands

Build, manage, and distribute Claude Code Skills with professional tooling.

Overview

The Skills namespace provides commands for creating comprehensive, production-ready Claude Code Skills through an elicitation-driven development process. Four specialized agents work together to ensure every skill follows best practices and meets quality standards.

Available Commands

/skills:build-skill

Primary command for creating new Claude Code Skills

Creates skills through a structured 4-phase process:

  1. Requirements Elicitation - Understand user needs through targeted questions
  2. Skill Generation - Build complete skill structure with files and code
  3. Validation & Testing - Comprehensive quality assurance and testing
  4. Documentation Enhancement - Professional documentation and examples

When to use:

  • Creating new skills from scratch
  • Converting workflows into reusable skills
  • Building team-shared capabilities
  • Packaging expertise for distribution

Example:

/skills:build-skill

"I need a skill for managing database migrations"
"Create a skill that helps with PDF form filling"
"Build a skill for writing API documentation"

The Skill Builder Agents

🎯 skill-elicitation-agent

Expert at gathering requirements and creating specifications

Responsibilities:

  • Ask 3-5 targeted questions to understand needs
  • Convert user requirements into detailed specifications
  • Plan skill structure and components
  • Validate completeness before generation

Triggers automatically when: Building new skills or gathering requirements

🏗️ skill-generator-agent

Expert at creating skill files and code

Responsibilities:

  • Generate SKILL.md with proper structure
  • Create supporting files (reference.md, examples.md)
  • Write scripts with error handling
  • Set up directory structure
  • Document dependencies

Triggers automatically when: Creating skills from specifications

✅ skill-validator-agent

Expert at testing and quality assurance

Responsibilities:

  • Validate YAML frontmatter
  • Test skill structure and organization
  • Execute and test scripts
  • Check description quality
  • Verify progressive disclosure
  • Generate validation reports

Triggers automatically when: Validating new or modified skills

📚 skill-documenter-agent

Expert at creating comprehensive documentation

Responsibilities:

  • Enhance SKILL.md with examples
  • Create detailed reference documentation
  • Generate example collections
  • Write troubleshooting guides
  • Ensure documentation quality

Triggers automatically when: Documenting skills

Skill Types

Simple Skills

Single SKILL.md file with instructions

Best for:

  • Instruction-based workflows
  • Process guidance
  • Best practices
  • Checklists

Example: Commit message formatting, code review guidelines

Multi-File Skills

SKILL.md + reference docs + examples

Best for:

  • Complex workflows
  • Multiple sub-tasks
  • Deep technical content
  • Progressive learning

Example: PDF processing, API design, database management

Tool-Restricted Skills

Skills with allowed-tools specified

Best for:

  • Read-only operations
  • Safety-critical workflows
  • Scoped capabilities
  • Security-focused tasks

Example: Security analysis, code auditing

Code-Execution Skills

Skills with bundled scripts

Best for:

  • Deterministic operations
  • Complex algorithms
  • Data processing
  • File manipulation

Example: Data transformation, file generation, validation

Skill Locations

Personal Skills: ~/.claude/skills/

Your private skills, not shared

✅ Use for:

  • Individual workflows and preferences
  • Experimental skills
  • Personal productivity tools
  • Private utilities

Project Skills: .claude/skills/

Team-shared, version-controlled

✅ Use for:

  • Team workflows and conventions
  • Project-specific expertise
  • Shared utilities
  • Collaborative tools

Plugin Skills: Plugin directory

Distributed via marketplace

✅ Use for:

  • Public sharing
  • Marketplace deployment
  • Bundled capabilities
  • Wide distribution

Best Practices

1. Start with Elicitation

  • Don't skip the questions phase
  • Understand requirements deeply
  • Get user approval on specification

2. Follow the Complete Flow

  • Use all four agents sequentially
  • Don't shortcut the process
  • Each agent adds critical value

3. Validate Thoroughly

  • Fix issues immediately
  • Re-validate after changes
  • Don't proceed with failures

4. Document Comprehensively

  • Clear instructions
  • Multiple examples
  • Troubleshooting guides
  • Best practices

5. Test Realistically

  • Use real-world scenarios
  • Test edge cases
  • Verify script execution
  • Check skill triggering

Workflow Examples

Example 1: Creating a Simple Skill

Goal: Skill for writing conventional commit messages

# Start the process
/skills:build-skill

# Elicitation phase
Q: What commit message format? (Conventional Commits)
Q: Any project-specific requirements? (None)
Q: Example commits to reference? (Yes, from CHANGELOG)

# Generation creates:
- commit-helper/SKILL.md (instructions + examples)

# Validation passes all checks

# Documentation adds:
- Best practices section
- Multiple examples
- Troubleshooting tips

# Result: Production-ready commit helper skill

Example 2: Multi-File Skill with Scripts

Goal: PDF form filling skill

# Start the process
/skills:build-skill

# Elicitation phase
Q: What PDF operations? (Fill forms, extract fields)
Q: Which libraries? (pypdf, pdfplumber)
Q: Safety restrictions? (None, can write files)

# Generation creates:
- pdf-form-filler/SKILL.md (core instructions)
- pdf-form-filler/FORMS.md (form-specific details)
- pdf-form-filler/REFERENCE.md (API docs)
- pdf-form-filler/scripts/fill_form.py
- pdf-form-filler/scripts/validate.py

# Validation tests:
- YAML structure ✅
- Script execution ✅
- Dependencies documented ✅

# Documentation adds:
- 10+ examples
- Troubleshooting section
- Integration patterns

# Result: Comprehensive PDF processing skill

Example 3: Tool-Restricted Read-Only Skill

Goal: Security code analysis skill

# Start the process
/skills:build-skill

# Elicitation phase
Q: Analysis scope? (OWASP Top 10 vulnerabilities)
Q: Should modify files? (No, read-only)
Q: Output format? (Detailed report)

# Generation creates:
- security-analyzer/SKILL.md
  - allowed-tools: Read, Grep, Glob
- security-analyzer/PATTERNS.md (vulnerability patterns)
- security-analyzer/REFERENCE.md (OWASP guide)

# Validation confirms:
- Tool restrictions correct ✅
- No write operations ✅
- Security patterns valid ✅

# Documentation adds:
- Security best practices
- Example vulnerabilities
- Remediation guides

# Result: Safe, read-only security skill

Quality Standards

Every skill created through /skills:build-skill meets these standards:

Structure

  • Valid YAML frontmatter
  • Clear, specific description
  • Proper file organization
  • Progressive disclosure

Content

  • Actionable instructions
  • Comprehensive examples
  • Best practices
  • Troubleshooting guides

Code

  • Syntax validation
  • Error handling
  • Security checks
  • Proper permissions

Documentation

  • Clear writing
  • Complete coverage
  • Tested examples
  • Usage guides

Validation

  • Passes all quality checks
  • Score 8/10 or higher
  • No critical issues
  • Production-ready

Common Use Cases

Team Onboarding

Create skills for:

  • Project setup and configuration
  • Team coding standards
  • Common workflows
  • Troubleshooting guides

Domain Expertise

Package knowledge about:

  • Specialized technologies
  • Industry-specific workflows
  • Complex algorithms
  • Best practices

Automation

Build skills that:

  • Execute repetitive tasks
  • Process data deterministically
  • Generate boilerplate
  • Validate outputs

Quality Assurance

Create tools for:

  • Code review checklists
  • Security analysis
  • Performance optimization
  • Testing workflows

Integration with Other Commands

Skills work seamlessly with other commands:

  • /project:* - Project skills auto-load for projects
  • /dev:* - Dev tools can leverage skill capabilities
  • /test:* - Skills can include testing workflows
  • /docs:* - Skills can generate documentation

Troubleshooting

Skill Not Loading

Check:

  • File location: ~/.claude/skills/ or .claude/skills/
  • YAML syntax: Valid frontmatter
  • Restart: Claude Code to reload skills

Skill Not Triggering

Fix:

  • Use explicit keywords from description
  • Try: "Use [skill-name] skill to [task]"
  • Check description includes trigger words

Validation Failures

Process:

  1. Review validation report
  2. Fix critical issues first
  3. Re-run skill-validator-agent
  4. Iterate until passing

Script Errors

Check:

  • Dependencies installed
  • File permissions (chmod +x)
  • Syntax validation
  • Error handling

Resources

Documentation

Examples

Support

  • Ask Claude: "How do I create a skill for [task]?"
  • Use /skills:build-skill for guided creation
  • Check validation reports for issues

Future Enhancements

Coming soon:

  • Skill testing framework
  • Skill marketplace integration
  • Version management
  • Dependency tracking
  • Usage analytics
  • Automated updates

Getting Started

Ready to build your first skill?

/skills:build-skill

The skill-elicitation-agent will guide you through the process!


Transform your workflows into reusable skills with professional tooling and quality assurance.