lite-skill-generator

name	lite-skill-generator
description	Lightweight skill generator with style learning - creates simple skills using flow-based execution and style imitation. Use for quick skill scaffolding, simple workflow creation, or style-aware skill generation.
allowed-tools	Read, Write, Bash, Glob, Grep, AskUserQuestion

Lite Skill Generator

Lightweight meta-skill for rapid skill creation with intelligent style learning and flow-based execution.

Core Concept

Simplicity First: Generate simple, focused skills quickly with minimal overhead. Learn from existing skills to maintain consistent style and structure.

Progressive Disclosure: Follow anthropics' three-layer loading principle:

1. Metadata - name, description, triggers (always loaded)
2. SKILL.md - core instructions (loaded when triggered)
3. Bundled resources - scripts, references, assets (loaded on demand)

Execution Model

3-Phase Flow: Style Learning → Requirements Gathering → Generation

User Input → Phase 1: Style Analysis → Phase 2: Requirements → Phase 3: Generate → Skill Package
                ↓                            ↓                        ↓
         Learn from examples      Interactive prompts      Write files + validate

Architecture Overview

┌─────────────────────────────────────────────────────────────────┐
│                    Lite Skill Generator                          │
│                                                                  │
│  Input: Skill name, purpose, reference skills                   │
│                         ↓                                        │
│  ┌─────────────────────────────────────────────────────────┐    │
│  │  Phase 1-3: Lightweight Pipeline                        │    │
│  │  ┌────┐ ┌────┐ ┌────┐                                  │    │
│  │  │ P1 │→│ P2 │→│ P3 │                                  │    │
│  │  │Styl│ │Req │ │Gen │                                  │    │
│  │  └────┘ └────┘ └────┘                                  │    │
│  └─────────────────────────────────────────────────────────┘    │
│                         ↓                                        │
│  Output: .claude/skills/{skill-name}/ (minimal package)         │
│                                                                  │
└─────────────────────────────────────────────────────────────────┘

3-Phase Workflow

Phase 1: Style Analysis & Learning

Analyze reference skills to extract language patterns, structural conventions, and writing style.

// Phase 1 Execution Flow
async function analyzeStyle(referencePaths) {
  // Step 1: Load reference skills
  const references = [];
  for (const path of referencePaths) {
    const content = Read(path);
    references.push({
      path: path,
      content: content,
      metadata: extractYAMLFrontmatter(content)
    });
  }

  // Step 2: Extract style patterns
  const styleProfile = {
    // Structural patterns
    structure: {
      hasFrontmatter: references.every(r => r.metadata !== null),
      sectionHeaders: extractCommonSections(references),
      codeBlockUsage: detectCodeBlockPatterns(references),
      flowDiagramUsage: detectFlowDiagrams(references)
    },

    // Language patterns
    language: {
      instructionStyle: detectInstructionStyle(references), // 'imperative' | 'declarative' | 'procedural'
      pseudocodeUsage: detectPseudocodePatterns(references),
      verbosity: calculateVerbosityLevel(references),      // 'concise' | 'detailed' | 'verbose'
      terminology: (references)
    },

    
    : {
      : (references),      
      : (references),
      : (references)
    }
  };

  
   {
    : styleProfile,
    : (styleProfile),
    : (references, styleProfile)
  };
}


 () {
   allSections = references.(
    r..()?.( s.(, ))
  ).();
   (allSections);
}


 () {
   imperativePattern = ;
   declarativePattern = ;
   proceduralPattern = ;

   scores = references.( ({
    : (r..(imperativePattern) || []).,
    : (r..(declarativePattern) || []).,
    : (r..(proceduralPattern) || []).
  }));

   (scores);
}


 () {
   hasJavaScriptBlocks = references.( r..());
   hasFunctionDefs = references.( .(r.));
   hasFlowComments = references.( .(r.));

   {
    : hasJavaScriptBlocks && hasFunctionDefs,
    : hasFlowComments,
    : hasFunctionDefs ?  : 
  };
}

Output:

Style Analysis Complete:
  Structure: Flow-based with pseudocode
  Language: Procedural, detailed
  Organization: Sequential phases
  Key Patterns: 3-5 phases, function definitions, ASCII diagrams

Recommendations:
  ✓ Use phase-based structure (3-4 phases)
  ✓ Include pseudocode for complex logic
  ✓ Add ASCII flow diagrams
  ✓ Maintain concise documentation style

---

Phase 2: Requirements Gathering

Interactive discovery of skill requirements using learned style patterns.

async function gatherRequirements(styleProfile) {
  // Step 1: Basic information
  const basicInfo = await AskUserQuestion({
    questions: [
      {
        question: "What is the skill name? (kebab-case, e.g., 'pdf-generator')",
        header: "Name",
        options: [
          { label: "pdf-generator", description: "Example: PDF generation skill" },
          { label: "code-analyzer", description: "Example: Code analysis skill" },
          { label: "Custom", description: "Enter custom name" }
        ]
      },
      {
        question: "What is the primary purpose?",
        header: "Purpose",
        options: [
          { label: "Generation", description: "Create/generate artifacts" },
          { label: "Analysis", description: "Analyze/inspect code or data" },
          { label: "Transformation", description: "Convert/transform content" },
          { label: "Orchestration", description: "Coordinate multiple operations" }
        ]
      }
    ]
  });

  
   complexity =  ({
    : [{
      : ,
      : ,
      : [
        { : , :  },
        { : , :  },
        { : , :  }
      ]
    }]
  });

  
   tools =  ({
    : [{
      : ,
      : ,
      : ,
      : [
        { : , :  },
        { : , :  },
        { : , :  },
        { : , :  },
        { : , :  }
      ]
    }]
  });

  
   output =  ({
    : [{
      : ,
      : ,
      : [
        { : , :  },
        { : , :  },
        { : , :  },
        { : , :  }
      ]
    }]
  });

  
   {
    : basicInfo.,
    : basicInfo.,
    : (basicInfo., basicInfo.),
    : (complexity.),
    : tools.,
    : output.,
    : styleProfile,
    : (basicInfo., basicInfo.)
  };
}


 () {
   templates = {
    : ,
    : ,
    : ,
    : 
  };
   templates[purpose] || ;
}


 () {
   base = [name, name.(, )];
   purposeVariants = {
    : [, , ],
    : [, , ],
    : [, , ],
    : [, , ]
  };
   [...base, ...purposeVariants[purpose].( )];
}

Display to User:

Requirements Gathered:
  Name: pdf-generator
  Purpose: Generation
  Steps: 3 (Setup → Generate → Validate)
  Tools: Read, Write, Bash
  Output: Single file (PDF document)
  Triggers: "pdf-generator", "generate pdf", "create pdf"

Style Application:
  Using flow-based structure (from style analysis)
  Including pseudocode blocks
  Adding ASCII diagrams for clarity

---

Phase 3: Generate Skill Package

Create minimal skill structure with style-aware content generation.

async function generateSkillPackage(requirements) {
  const skillDir = `.claude/skills/${requirements.name}`;
  const workDir = `.workflow/.scratchpad/lite-skill-gen-${Date.now()}`;

  // Step 1: Create directory structure
  Bash(`mkdir -p "${skillDir}" "${workDir}"`);

  // Step 2: Generate SKILL.md (using learned style)
  const skillContent = generateSkillMd(requirements);
  Write(`${skillDir}/SKILL.md`, skillContent);

  // Step 3: Conditionally add bundled resources
  if (requirements.outputType === 'Directory structure') {
    Bash(`mkdir -p "${skillDir}/templates"`);
    const templateContent = generateTemplate(requirements);
    Write(`${skillDir}/templates/base-template.md`, templateContent);
  }

  if (requirements.allowedTools.includes('Bash')) {
    Bash(`mkdir -p "${skillDir}/scripts"`);
    const scriptContent = generateScript(requirements);
    Write(`${skillDir}/scripts/helper.sh`, scriptContent);
  }

  
   readmeContent = (requirements);
  (, readmeContent);

  
   validation = (skillDir, requirements);
  (, .(validation, , ));

  
   {
    : skillDir,
    : [
      ,
      ...(validation. ? [] : []),
      ...(validation. ? [] : []),
      
    ],
    : validation,
    : (requirements)
  };
}


 () {
   { styleProfile } = req;

  
   frontmatter = ;

  
   content = frontmatter;

  content += ;
  content += ;

  
   (styleProfile..) {
    content += (req);
  }

  
  content += ;
   (styleProfile...) {
    content += (req);
  }  {
    content += (req);
  }

  
   ( i = ; i < req.; i++) {
    content += (i + , req, styleProfile);
  }

  
   (styleProfile.. !== ) {
    content += (req);
  }

   content;
}


 () {
   ;
}


 () {
   ;
}


 () {
   phaseName = (phaseNum, req);

   section = ;

   (styleProfile...) {
    section += ;
    section += ;
    section += ;
    section += ;
    section += ;
  }  {
    section += ;
    section += ;
    section += ;
    section += ;
  }

   section;
}


 () {
   requiredFiles = [, ];
   exists = requiredFiles.( (). === );

   {
    : exists.( e),
    : (). === ,
    : (). === ,
    : requiredFiles.( exists[i]),
    : (skillDir, req.)
  };
}

Output:

Skill Package Generated:
  Location: .claude/skills/pdf-generator/

Structure:
  ✓ SKILL.md (entry point)
  ✓ README.md (usage guide)
  ✓ templates/ (directory templates)
  ✓ scripts/ (helper scripts)

Validation:
  ✓ All required files present
  ✓ Style compliance: 95%
  ✓ Frontmatter valid
  ✓ Tool references correct

Next Steps:
  1. Review SKILL.md and customize phases
  2. Test skill: /skill:pdf-generator "test input"
  3. Iterate based on usage

---

Complete Execution Flow

User: "Create a PDF generator skill"
    ↓
Phase 1: Style Analysis
    |-- Read reference skills (ccw.md, ccw-coordinator.md)
    |-- Extract style patterns (flow diagrams, pseudocode, structure)
    |-- Generate style profile
    +-- Output: Style recommendations
    ↓
Phase 2: Requirements
    |-- Ask: Name, purpose, steps
    |-- Ask: Tools, output format
    |-- Generate: Description, triggers
    +-- Output: Requirements config
    ↓
Phase 3: Generation
    |-- Create: Directory structure
    |-- Write: SKILL.md (style-aware)
    |-- Write: README.md
    |-- Optionally: templates/, scripts/
    |-- Validate: Structure and style
    +-- Output: Skill package
    ↓
Return: Skill location + next steps

Phase Execution Protocol

// Main entry point
async function liteSkillGenerator(input) {
  // Phase 1: Style Learning
  const references = [
    '.claude/commands/ccw.md',
    '.claude/commands/ccw-coordinator.md',
    ...discoverReferenceSkills(input)
  ];
  const styleProfile = await analyzeStyle(references);
  console.log(`Style Analysis: ${styleProfile.organization.phaseStructure}, ${styleProfile.language.verbosity}`);

  // Phase 2: Requirements
  const requirements = await gatherRequirements(styleProfile);
  console.log(`Requirements: ${requirements.name} (${requirements.steps} phases)`);

  // Phase 3: Generation
  const result = await generateSkillPackage(requirements);
  console.log(`✅ Generated: ${result.skillPath}`);

  return result;
}

Output Structure

Minimal Package (default):

.claude/skills/{skill-name}/
├── SKILL.md                # Entry point with frontmatter
└── README.md               # Usage guide

With Templates (if needed):

.claude/skills/{skill-name}/
├── SKILL.md
├── README.md
└── templates/
    └── base-template.md

With Scripts (if using Bash):

.claude/skills/{skill-name}/
├── SKILL.md
├── README.md
└── scripts/
    └── helper.sh

Key Design Principles

1. Style Learning - Analyze reference skills to maintain consistency
2. Minimal Overhead - Generate only essential files (SKILL.md + README)
3. Progressive Disclosure - Follow anthropics' three-layer loading
4. Flow-Based - Use pseudocode and flow diagrams (when style appropriate)
5. Interactive - Guided requirements gathering via AskUserQuestion
6. Fast Generation - 3 phases instead of 6, focused on simplicity
7. Style Awareness - Adapt output based on detected patterns

Style Pattern Detection

Structural Patterns:

• YAML frontmatter usage (100% in references)
• Section headers (H2 for major, H3 for sub-sections)
• Code blocks (JavaScript pseudocode, Bash examples)
• ASCII diagrams (architecture, flow charts)

Language Patterns:

• Instruction style: Procedural with function definitions
• Pseudocode: JavaScript-based with flow annotations
• Verbosity: Detailed but focused
• Terminology: Phase, workflow, pipeline, orchestrator

Organization Patterns:

• Phase structure: 3-5 sequential phases
• Example density: Moderate (1-2 per major section)
• Template usage: Minimal (only when necessary)

Usage Examples

Basic Generation:

User: "Create a markdown formatter skill"
Lite-Skill-Generator:
  → Analyzes ccw.md style
  → Asks: Name? "markdown-formatter"
  → Asks: Purpose? "Transformation"
  → Asks: Steps? "3 steps"
  → Generates: .claude/skills/markdown-formatter/

With Custom References:

User: "Create a skill like software-manual but simpler"
Lite-Skill-Generator:
  → Analyzes software-manual skill
  → Learns: Multi-phase, agent-based, template-heavy
  → Simplifies: 3 phases, direct execution, minimal templates
  → Generates: Simplified version

Comparison: lite-skill-generator vs skill-generator

Aspect	lite-skill-generator	skill-generator
Phases	3 (Style → Req → Gen)	6 (Spec → Req → Dir → Gen → Specs → Val)
Style Learning	Yes (analyze references)	No (fixed templates)
Complexity	Simple skills only	Full-featured skills
Output	Minimal (SKILL.md + README)	Complete (phases/, specs/, templates/)
Generation Time	Fast (~2 min)	Thorough (~10 min)
Use Case	Quick scaffolding	Production-ready skills

Workflow Integration

Standalone:

/skill:lite-skill-generator "Create a log analyzer skill"

With References:

/skill:lite-skill-generator "Create a skill based on ccw-coordinator.md style"

Batch Generation (for multiple simple skills):

/skill:lite-skill-generator "Create 3 skills: json-validator, yaml-parser, toml-converter"

---

Next Steps After Generation:

1. Review .claude/skills/{name}/SKILL.md
2. Customize phase logic for your use case
3. Add examples to README.md
4. Test skill with sample input
5. Iterate based on real usage