skills-collection-manager

Skills Collection Manager

Efficiently manage, organize, and maintain large collections of Claude Code skills at scale.

Overview

As skill collections grow to hundreds of entries, manual management becomes impractical. This skill provides automated tools and workflows for:

• Bulk downloading skills from GitHub repositories
• Organizing skills into logical categories
• Detecting and removing duplicates
• Consolidating similar skills
• Maintaining repository health
• Generating documentation and indexes

When to Use

Use this skill when:

• Managing 50+ skills in a collection
• Importing skills from multiple GitHub repositories
• Organizing an unstructured skill directory
• Identifying duplicate or redundant skills
• Creating team-wide skill libraries
• Maintaining organizational skill repositories
• Building curated skill collections
• Auditing skill quality and coverage

Directory Structure Patterns

Pattern 1: Flat Structure

skills/
├── docker-helper/
│   └── SKILL.md
├── kubernetes-deploy/
│   └── SKILL.md
├── postgres-optimization/
│   └── SKILL.md
└── ...

Pros: Simple, easy to navigate Cons: Hard to manage at scale (100+ skills)

Pattern 2: Categorized Structure

skills/
├── automation/
│   ├── skill-harvester/
│   │   └── SKILL.md
│   └── workflow-automation/
│       └── SKILL.md
├── backend/
│   ├── api-design/
│   │   └── SKILL.md
│   └── database-optimization/
│       └── SKILL.md
├── devops/
│   ├── docker-optimization/
│   │   └── SKILL.md
│   └── kubernetes-deploy/
│       └── SKILL.md
└── ...

Pros: Organized, scalable Cons: Requires categorization logic

Pattern 3: Hybrid Structure

skills/              # Flat for easy access
skills-by-category/  # Organized for browsing
duplicates/          # Quarantine area
archived/            # Old/deprecated skills

Pros: Best of both worlds Cons: More complex to maintain

Bulk Skill Download

Download from GitHub Repository

#!/bin/bash
# bulk-download-skills.sh - Download skills from GitHub repos

REPOS_FILE="${1:-repos.txt}"
OUTPUT_DIR="downloaded-skills"
CHECKPOINT="download-checkpoint.txt"

# Create repos list if it doesn't exist
if [ ! -f "$REPOS_FILE" ]; then
    cat > "$REPOS_FILE" << EOF
https://github.com/anthropics/claude-code-skills
https://github.com/user/custom-skills
https://github.com/team/shared-skills
EOF
    echo "Created example $REPOS_FILE - edit and run again"
    exit 0
fi

mkdir -p "$OUTPUT_DIR"

# Process repos in batches
while read -r repo_url; do
    # Skip comments and empty lines
    [[ "$repo_url" =~ ^# ]] && continue
    [[ -z "$repo_url" ]] && continue

    # Check if already downloaded
    if grep -q "$repo_url" "$CHECKPOINT" 2>/dev/null; then
        echo "✓ Skipping $repo_url (already downloaded)"
        continue
    fi

    echo "=== Downloading: $repo_url ==="

    # Extract repo name
    REPO_NAME=$(basename "$repo_url" .git)
    TEMP_DIR=$(mktemp -d)

    # Clone with timeout
    if timeout 60s git clone --depth 1 "$repo_url" "$TEMP_DIR" 2>/dev/null; then
        # Find and copy skill files
        SKILLS_FOUND=0

        # Look for skills in common locations
        for PATTERN in "skills/*/SKILL.md" "skills/**/skill.md" ".claude/skills/*/SKILL.md"; do
            find "$TEMP_DIR" -path "*/$PATTERN" 2>/dev/null | while read skill_file; do
                # Extract skill name (parent directory)
                SKILL_NAME=$(basename $(dirname "$skill_file"))
                DEST_DIR="$OUTPUT_DIR/${REPO_NAME}/${SKILL_NAME}"

                mkdir -p "$DEST_DIR"
                cp "$skill_file" "$DEST_DIR/"

                # Copy additional files if present
                SKILL_DIR=$(dirname "$skill_file")
                cp "$SKILL_DIR"/*.md "$DEST_DIR/" 2>/dev/null || true
                cp "$SKILL_DIR"/*.yaml "$DEST_DIR/" 2>/dev/null || true
                cp "$SKILL_DIR"/*.json "$DEST_DIR/" 2>/dev/null || true

                SKILLS_FOUND=$((SKILLS_FOUND + 1))
                echo "  ✓ Copied: $SKILL_NAME"
            done
        done

        if [ $SKILLS_FOUND -gt 0 ]; then
            echo "$repo_url" >> "$CHECKPOINT"
            echo "✓ Downloaded $SKILLS_FOUND skills from $REPO_NAME"
        else
            echo "⚠️  No skills found in $REPO_NAME"
        fi

        rm -rf "$TEMP_DIR"
    else
        echo "✗ Failed to clone $repo_url"
    fi

    echo ""
done < "$REPOS_FILE"

# Summary
TOTAL_SKILLS=$(find "$OUTPUT_DIR" -name "SKILL.md" -o -name "skill.md" | wc -l)
echo "=== Download Complete ==="
echo "Total skills downloaded: $TOTAL_SKILLS"
echo "Output directory: $OUTPUT_DIR"

Batch Download Script

#!/bin/bash
# download-top-repos.sh - Download skills from popular repositories

# Top Claude Code skill repositories
REPOS=(
    "https://github.com/anthropics/claude-code-skills"
    "https://github.com/works/claude-code-skills"
    "https://github.com/jonbaker99/my-claude-code-skills"
    "https://github.com/diet103/my_claude_skills"
    "https://github.com/mrgoonie/my-claude-code-skills"
)

BATCH_SIZE=5
PROCESSED=0

for repo in "${REPOS[@]}"; do
    echo "$repo" >> repos-batch.txt
    PROCESSED=$((PROCESSED + 1))

    # Process in batches to avoid timeouts
    if [ $((PROCESSED % BATCH_SIZE)) -eq 0 ]; then
        ./bulk-download-skills.sh repos-batch.txt
        rm repos-batch.txt
        echo "Batch complete. Continuing..."
        sleep 2
    fi
done

# Process remaining
if [ -f repos-batch.txt ]; then
    ./bulk-download-skills.sh repos-batch.txt
    rm repos-batch.txt
fi

Skill Organization

Auto-Categorization

#!/bin/bash
# categorize-skills.sh - Auto-categorize skills based on content

SKILLS_DIR="${1:-skills}"
OUTPUT_DIR="skills-by-category"

# Category mapping based on keywords
declare -A CATEGORIES=(
    ["docker|container|kubernetes|k8s"]="infrastructure"
    ["api|endpoint|rest|graphql|backend"]="backend"
    ["react|vue|angular|frontend|ui|component"]="frontend"
    ["test|testing|jest|pytest|mocha"]="testing"
    ["ci|cd|deploy|github.action|jenkins"]="devops"
    ["database|postgres|mysql|mongodb|sql"]="databases"
    ["auth|authentication|jwt|oauth|security"]="security"
    ["aws|azure|gcp|cloud|serverless"]="cloud"
    ["python|javascript|typescript|golang|rust"]="development"
    ["doc|documentation|readme|markdown"]="documentation"
)

mkdir -p "$OUTPUT_DIR"

# Process each skill
find "$SKILLS_DIR" -name "SKILL.md" -o -name "skill.md" | while read skill_file; do
    SKILL_NAME=$(basename $(dirname "$skill_file"))
    SKILL_CONTENT=$(cat "$skill_file" | tr '[:upper:]' '[:lower:]')

    echo "Processing: $SKILL_NAME"

    # Try to match category
    MATCHED_CATEGORY=""
    for pattern in "${!CATEGORIES[@]}"; do
        if echo "$SKILL_CONTENT" | grep -qE "$pattern"; then
            MATCHED_CATEGORY="${CATEGORIES[$pattern]}"
            break
        fi
    done

    # Default category if no match
    if [ -z "$MATCHED_CATEGORY" ]; then
        MATCHED_CATEGORY="uncategorized"
    fi

    # Copy to categorized directory
    DEST_DIR="$OUTPUT_DIR/$MATCHED_CATEGORY/$SKILL_NAME"
    mkdir -p "$DEST_DIR"
    cp -r "$(dirname $skill_file)"/* "$DEST_DIR/"

    echo "  → $MATCHED_CATEGORY"
done

# Generate summary
echo ""
echo "=== Categorization Summary ==="
for category_dir in "$OUTPUT_DIR"/*; do
    CATEGORY=$(basename "$category_dir")
    COUNT=$(find "$category_dir" -name "SKILL.md" -o -name "skill.md" | wc -l)
    printf "%-20s : %3d skills\n" "$CATEGORY" "$COUNT"
done

Manual Reorganization

#!/bin/bash
# reorganize-skills.sh - Interactive skill reorganization

SKILLS_DIR="skills"
CATEGORIES=("automation" "backend" "cloud" "data-engineering" "devops" "documentation" "frontend" "infrastructure" "security" "testing")

# Show uncategorized skills
echo "=== Uncategorized Skills ==="
find "$SKILLS_DIR" -maxdepth 1 -type d | tail -n +2 | while read skill_dir; do
    SKILL_NAME=$(basename "$skill