npm package discovery and stats viewer.

Discover Tips

  • General search

    [free text search, go nuts!]

  • Package details

    pkg:[package-name]

  • User packages

    @[username]

Sponsor

Optimize Toolset

I’ve always been into building performant and accessible sites, but lately I’ve been taking it extremely seriously. So much so that I’ve been building a tool to help me optimize and monitor the sites that I build to make sure that I’m making an attempt to offer the best experience to those who visit them. If you’re into performant, accessible and SEO friendly sites, you might like it too! You can check it out at Optimize Toolset.

About

Hi, 👋, I’m Ryan Hefner  and I built this site for me, and you! The goal of this site was to provide an easy way for me to check the stats on my npm packages, both for prioritizing issues and updates, and to give me a little kick in the pants to keep up on stuff.

As I was building it, I realized that I was actually using the tool to build the tool, and figured I might as well put this out there and hopefully others will find it to be a fast and useful way to search and browse npm packages as I have.

If you’re interested in other things I’m working on, follow me on Twitter or check out the open source projects I’ve been publishing on GitHub.

I am also working on a Twitter bot for this site to tweet the most popular, newest, random packages from npm. Please follow that account now and it will start sending out packages soon–ish.

Open Software & Tools

This site wouldn’t be possible without the immense generosity and tireless efforts from the people who make contributions to the world and share their work via open source initiatives. Thank you 🙏

© 2026 – Pkg Stats / Ryan Hefner

@thuanha-groove/gck

v0.0.5

Published

CLI tool for bootstrapping and updating ClaudeKit projects

Readme

ClaudeKit Config UI

Command-line tool and web dashboard for managing ClaudeKit projects.

Version: 1.17.0

Overview

ClaudeKit Config UI (gck) provides both CLI and web dashboard for managing ClaudeKit projects. It is built with Bun, TypeScript, and React for development, while the published CLI runs on plain Node.js so end users do not need Bun installed.

Key Features:

  • CLI Commands (16): new, init, config, projects, setup, skills, agents, commands, migrate, doctor, versions, update, uninstall, watch, content, easter-egg
  • Web Dashboard: Interactive React UI via gck config ui for configuration and project management
  • Desktop App: The same dashboard as an installable macOS/Windows app, with every gck command runnable from the UI (see Desktop App)
  • Hook Diagnostics Dashboard: Inspect recent Claude hook activity and failures from gck config across global and project scopes
  • Projects Registry: Centralized registry at ~/.claudekit/projects.json with file locking
  • Skill Installation: Install ClaudeKit skills to other coding agents (Cursor, Codex, etc.)
  • Multi-tier Authentication: gh CLI → env vars → keychain → prompt fallback
  • Smart Merging: Conflict detection with user customization preservation
  • Skills Migration: Auto-detects and migrates skills structure changes
  • Offline Installation: From local archives or directories
  • Security: Path traversal protection, symlink validation, UNC path protection
  • Cross-Platform: macOS, Linux, Windows with platform-specific optimizations
  • Update Notifications: Intelligent 7-day cache for version checks

Documentation

Comprehensive documentation in /docs:

Prerequisites

Before using ClaudeKit CLI, you need to:

  1. Purchase a ClaudeKit Starter Kit from ClaudeKit.cc
  2. Get Repository Access: After purchase, you'll receive access to the private GitHub repository containing your kit
  3. Create a GitHub Personal Access Token (PAT) with repo scope to download releases

Without a purchased kit and repository access, the CLI will not be able to download any project templates.

Installation

The ClaudeKit CLI is published on npm at npmjs.com/package/gck.

End-user runtime note: global installs from npm, pnpm, yarn, or bun all execute the packaged Node.js CLI. Bun is optional for users and only needed for local ClaudeKit CLI development workflows. Most commands do not require native build tools; gck content uses an optional SQLite driver and shows remediation steps if that driver is unavailable.

Using npm (Recommended)

npm install -g gck

Using Bun

bun add -g gck

Using Yarn

yarn global add gck

Using pnpm

pnpm add -g gck

After installation, verify it's working:

gck --version

Desktop App (no Node.js required)

Prefer a window over a terminal? Download GCK for macOS or Windows from github.com/thuanha-groove/gck-release/releases/latest. The app bundles the dashboard and the CLI, shares the same config files as a terminal gck, and tells you when a new version is available. Builds are not code-signed yet, so the first launch needs Open anyway on macOS or Run anyway on Windows. Details in docs/desktop-app.md.

Usage

Discoverability Quick Start

# Top-level command discovery
gck --help

# Open config dashboard immediately
gck config

# Expose the dashboard intentionally to your LAN/Tailscale
gck config --host 0.0.0.0 --no-open

# Command-level help (recommended)
gck config --help
gck skills --help
gck agents --help
gck commands --help
gck migrate --help

Config Dashboard Access

By default, gck config binds the dashboard to 127.0.0.1 for local-only access.

Use --host when you intentionally want remote access from another device on the same trusted network:

# Bind to all interfaces
gck config --host 0.0.0.0 --no-open

# Bind to a specific interface or hostname
gck config --host 100.88.12.4 --no-open
gck config --host dashboard.local --no-open

The dashboard still enforces same-origin browser access. Remote access works when you open the UI from the same host/origin that reaches the server, instead of relying on a hardcoded IP allowlist.

Create New Project

# Interactive mode
gck new

# With options
gck new --dir my-project --kit engineer

# Show beta versions
gck new --beta

# With exclude patterns
gck new --exclude "*.log" --exclude "temp/**"

# Optional packages (OpenCode, Gemini)
gck new --opencode --gemini

# Install skills dependencies (Python, Node packages, system tools)
gck new --install-skills

# Command prefix (/gck: namespace to avoid conflicts)
gck new --prefix

# Offline installation (from local archive or directory)
gck new --archive ~/downloads/engineer-v1.16.0.zip
gck new --kit-path ~/extracted-kit/

# Direct repo downloads are also supported
gck new --archive ~/downloads/claudekit-engineer-main.zip
gck new --kit-path ~/downloads/claudekit-engineer-main/

Flags:

  • --install-skills: Auto-install Python packages, system tools (FFmpeg, ImageMagick), Node.js packages
  • --prefix: Move commands to /gck: namespace (/plan → /gck:plan)
  • --beta: Show pre-release versions in selection
  • --opencode/--gemini: Install optional packages
  • --archive <path>: Use local archive (zip/tar.gz) instead of downloading
  • --kit-path <path>: Use local kit directory instead of downloading

--archive and --kit-path both accept direct repo downloads with a single wrapper directory, including GitHub "Download ZIP" archives and extracted repo folders that still contain claudekit-engineer-main/ or similar at the top level.

Initialize or Update Project

Note: Run from project root.

# Interactive mode
gck init

# Non-interactive mode with sensible defaults
gck init --yes
gck init -y

# Combine with other flags
gck init -g --kit engineer -y

# With options
gck init --kit engineer --beta

# Global mode (platform-specific paths)
gck init --global

# Fresh installation (⚠️ DESTRUCTIVE - removes ALL customizations)
gck init --fresh

# With exclude patterns and prefix
gck init --exclude "*.local" --prefix

# Offline installation (from local archive or directory)
gck init --archive ~/downloads/engineer-v1.16.0.zip
gck init --kit-path ~/extracted-kit/

# Direct repo downloads are also supported
gck init --archive ~/downloads/claudekit-engineer-main.zip
gck init --kit-path ~/downloads/claudekit-engineer-main/

Flags:

  • --yes/-y: Non-interactive mode with sensible defaults (skip all prompts)
  • --global/-g: Use platform-specific config (macOS/Linux: ~/.claude, Windows: %USERPROFILE%.claude)
  • --fresh: Clean reinstall, removes .claude directory (requires "yes" confirmation)
  • --beta: Show pre-release versions
  • --prefix: Apply /gck: namespace to commands
  • --archive <path>: Use local archive (zip/tar.gz) instead of downloading
  • --kit-path <path>: Use local kit directory instead of downloading

--archive and --kit-path both accept direct repo downloads with a single wrapper directory, including GitHub "Download ZIP" archives and extracted repo folders that still contain claudekit-engineer-main/ or similar at the top level.

Default Behavior with -y Flag:

| Prompt | Default | |--------|---------| | Select ClaudeKit | engineer (first option) | | Target directory | Current directory (.) | | Version selection | Latest stable release | | Google Gemini setup | Skip | | Other optional features | Skip |

Update CLI

Keep the ClaudeKit CLI up to date:

# Check for CLI updates
gck update --check

# Update to latest version
gck update

# Update to specific version
gck update --version 1.17.0

# Update to beta / skip confirmation
gck update --beta
gck update --yes

The CLI notifies you when updates are available via gck --version.

Skills Migration:

  • Auto-detects structure changes (flat → categorized)
  • Preserves customizations (SHA-256 hashing)
  • Creates backup before migration
  • Rollback on failure

List Available Versions

# Show all available versions for all kits
gck versions

# Filter by specific kit
gck versions --kit engineer
gck versions --kit marketing

# Show more versions (default: 30)
gck versions --limit 50

# Include prereleases and drafts
gck versions --all

Diagnostics & Doctor

# Full health check (default)
gck doctor

# Verbose mode with execution timing and command details
gck doctor --verbose

# Generate shareable diagnostic report (prompts for gist upload)
gck doctor --report

# Auto-fix all fixable issues
gck doctor --fix

# CI mode: no prompts, exit 1 on failures
gck doctor --check-only

# Machine-readable JSON output
gck doctor --json

# Combine flags
gck doctor --verbose --check-only --json
gck doctor --verbose --fix

Health Checks:

  • System: Node.js, npm, Python, pip, Claude CLI, git, gh CLI
  • ClaudeKit: Global/project installation, versions, skills, skill listing budget, duplicate skill inventory
  • Auth: GitHub CLI authentication, repository access
  • Project: package.json, node_modules, lock files
  • Modules: Dynamic skill dependency resolution

Auto-Fix Capabilities: | Issue | Fix Action | |-------|------------| | Missing dependencies | Install via package manager | | Missing gh auth | Run gh auth login | | Corrupted node_modules | Reinstall dependencies | | Missing global install | Run gck init --global | | Missing skill deps | Install in skill directory | | Missing/low Engineer skill listing budget | Ensure computed skillListingBudgetFraction and a valid ClaudeKit-recommended skillListingMaxDescChars ceiling in project settings |

gck doctor never writes skillOverrides, hides skills, or deletes duplicate skills automatically. It reports existing skillOverrides and inventory issues so all active project/global skills can stay user-invocable while listing pressure is managed through 200k-context-floor project settings and bounded descriptions.

Exit Codes:

  • 0: All checks pass or issues fixed
  • 1: Failures detected (only with --check-only)

Note: gck diagnose is deprecated. Use gck doctor instead.

Uninstall

Remove ClaudeKit installations from your system:

gck uninstall              # Interactive mode - prompts for scope and confirmation
gck uninstall --local      # Uninstall only local installation (current project)
gck uninstall --global     # Uninstall only global installation (~/.claude/)
gck uninstall -g --kit marketing  # Remove only global Marketing kit
gck uninstall -l -y        # Local only, skip confirmation
gck uninstall -g -y        # Global only, skip confirmation
gck uninstall --yes        # Non-interactive - skip confirmation (for scripts)

Scope Selection:

  • When both local and global installations exist, you'll be prompted to choose:
    • Local only: Remove from current project (.claude/)
    • Global only: Remove from user directory (~/.claude/)
    • Both: Remove all ClaudeKit installations
  • When multiple kits are installed in the selected scope, interactive uninstall prompts for:
    • Marketing kit only or Engineer kit only: Remove one kit and preserve the other
    • All ClaudeKit kits: Remove the full selected installation scope
  • Use --local or --global flags to skip the prompt
  • Use --kit engineer or --kit marketing to skip the kit prompt

What it does:

  • Detects local .claude directory in current project
  • Detects global ~/.claude ClaudeKit installation
  • Shows paths before deletion
  • Requires confirmation (unless --yes flag)
  • Removes ClaudeKit subdirectories (commands/, agents/, skills/, workflows/, hooks/, metadata.json)
  • Preserves user configs like settings.json, settings.local.json, and CLAUDE.md

Note: Only removes valid ClaudeKit installations (with metadata.json). Regular .claude directories from Claude Desktop are not affected.

Watch GitHub Issues (gck watch)

Autonomous daemon that monitors GitHub issues, analyzes them with Claude, generates plans, and creates PRs.

# Start watching (single repo)
gck watch

# Dry-run mode (no posts/PRs)
gck watch --dry-run

# Custom poll interval (ms)
gck watch --interval 60000

# Force restart (clear state)
gck watch --force

# Verbose logging
gck watch --verbose

Features: issue lifecycle management (10 statuses), Claude-powered brainstorming/planning, automatic PR creation, rate limiting (persisted across restarts), maintainer reply filtering, processedIssues TTL, optional git worktree isolation per issue, multi-repo support, graceful shutdown.

Config: .ck.json under watch key. See docs/ck-watch.md for full configuration reference.

Content Generation (gck content)

Daemon that scans git activity (commits, PRs, tags), generates social media content with Claude, and publishes to X/Twitter and Facebook.

# Interactive setup wizard
gck content setup

# Start daemon
gck content start

# Check status
gck content status

# View logs
gck content logs

# Queue manual content
gck content queue

# Review workflow
gck content approve <id>
gck content reject <id>

# Dry-run / verbose
gck content start --dry-run
gck content start --verbose

Features: 11-phase pipeline (scan → filter → classify → context → create → validate → review → photo → publish → engage → analyze), noise filtering, context caching (24h TTL), content validation, photo generation, 3 review modes (auto/manual/hybrid), quiet hours scheduling, engagement tracking, SQLite database, platform-specific adapters.

Runtime dependency: gck content uses the optional native SQLite driver better-sqlite3. If the driver is unavailable, other CLI commands still work and gck content prints reinstall/build-tool guidance.

Config: .ck.json under content key. See docs/ck-content.md for full configuration reference.

Other Commands

# Show CLI version (shows local + global kit versions)
gck --version

# Show help
gck --help
gck -h

# Command-specific help
gck new --help
gck init --help
gck config --help
gck skills --help
gck versions --help

Debugging

gck new --verbose              # Enable verbose logging
gck new --verbose --log-file debug.log  # Save to file
CLAUDEKIT_VERBOSE=1 gck new   # Via environment variable

Cache Configuration

Release data is cached locally to improve performance. You can configure the cache TTL:

# Set custom cache TTL (in seconds, default: 3600 = 1 hour)
CK_CACHE_TTL=7200 gck versions    # Cache for 2 hours
CK_CACHE_TTL=0 gck versions       # Disable caching (always fetch fresh)

# Permanent configuration (add to ~/.bashrc or ~/.zshrc)
export CK_CACHE_TTL=1800         # 30 minutes

Cache Location: ~/.claudekit/cache/releases/

Update Notifications

The gck --version command checks for newer versions of your installed ClaudeKit and displays a notification if an update is available. The check is cached for 7 days to minimize API calls.

Disable Update Notifications:

# Set environment variable to disable
NO_UPDATE_NOTIFIER=1 gck --version

# Windows (permanent)
[System.Environment]::SetEnvironmentVariable("NO_UPDATE_NOTIFIER", "1", [System.EnvironmentVariableTarget]::User)

# macOS/Linux (add to ~/.bashrc or ~/.zshrc)
export NO_UPDATE_NOTIFIER=1

Cache Location: ~/.claudekit/cache/version-check.json (Windows: %USERPROFILE%\.claudekit\cache\)

Authentication

The CLI requires GitHub authentication to download releases from private repositories.

Authentication Flow

┌─────────────────────────────────────────────────┐
│          Multi-Tier Authentication               │
│                                                  │
│  1. GitHub CLI (gh auth token)                  │
│       ↓ (if not available)                       │
│  2. Environment Variables (GITHUB_TOKEN)        │
│       ↓ (if not set)                             │
│  3. Config File (~/.claudekit/config.json)      │
│       ↓ (if not found)                           │
│  4. OS Keychain (secure storage)                │
│       ↓ (if not stored)                          │
│  5. User Prompt (with save option)              │
└─────────────────────────────────────────────────┘

Quick Setup

Step 1: Install GitHub CLI

# Windows
winget install GitHub.cli

# macOS
brew install gh

# Linux
sudo apt install gh

Step 2: Authenticate with GitHub CLI

gh auth login

When prompted, follow these steps:

  1. Select GitHub.com
  2. Select HTTPS (or SSH if preferred)
  3. Authenticate Git? → Yes
  4. Select Login with a web browser (⚠️ recommended)
  5. Copy the one-time code shown
  6. Press Enter to open browser and paste the code
  7. Authorize GitHub CLI

⚠️ Important: Select "Login with a web browser" - do NOT use "Paste an authentication token" as PAT authentication is no longer supported for accessing private repositories.

Troubleshooting

Run the doctor command to diagnose issues:

# Interactive diagnostics
gck doctor

# Generate report for support
gck doctor --report

# CI/automation
gck doctor --check-only --json

# Verbose logging
gck new --verbose
gck init --verbose

Common Issues:

  • "Access denied": Run gck doctor to check auth, use --fix to auto-repair
  • "Authentication failed": Run gck doctor --fix to re-authenticate, or manually run gh auth login (select 'Login with a web browser')
  • "GitHub CLI not authenticated": Run gh auth login and select 'Login with a web browser' (NOT 'Paste token')
  • Module errors: Run gck doctor --fix to reinstall skill dependencies
  • Need help: Run gck doctor --report and share the gist URL

Available Kits

ClaudeKit offers premium starter kits available for purchase at ClaudeKit.cc:

  • engineer: ClaudeKit Engineer - Engineering toolkit for building with Claude (v1.0.0+)
  • marketing: ClaudeKit Marketing - Content automation toolkit (v1.0.0 available)

Each kit provides a comprehensive project template with best practices, tooling, and workflows optimized for Claude Code development.

Configuration

Configuration is stored in ~/.claudekit/config.json:

{
  "github": {
    "token": "stored_in_keychain"
  },
  "defaults": {
    "kit": "engineer",
    "dir": "."
  }
}

Protected Files

The following file patterns are protected and will not be overwritten during updates:

  • .env, .env.local, .env.*.local
  • *.key, *.pem, *.p12
  • node_modules/**, .git/**
  • dist/**, build/**

Excluding Files

Use --exclude flag with glob patterns to skip files:

gck new --exclude "*.log" --exclude "temp/**"
gck update --exclude "node_modules/**" --exclude "dist/**"

Patterns: * (any chars), ** (recursive), ? (single char), [abc], {a,b} Restrictions: No absolute paths, no path traversal (..), 1-500 chars Note: User patterns are ADDED to default protected patterns

Custom .claude Files & Skills Migration

Custom File Preservation: The CLI automatically preserves your custom .claude/ files during updates:

  • Custom slash commands
  • Personal workflows
  • Project-specific configurations
  • Any other custom files in .claude/ directory

Skills Directory Migration: Automatic migration when structure changes (flat → categorized):

  • Detection: Manifest-based + heuristic fallback
  • Customizations: SHA-256 hash comparison detects modifications
  • Safety: Backup before migration, rollback on failure
  • Preservation: All customizations preserved during migration
  • Interactive: Prompts for confirmation (can skip in CI/CD)

Example Migration:

Before (flat):
  .claude/skills/
    ├── gemini-vision/
    ├── postgresql-psql/
    └── cloudflare-dns/

After (categorized):
  .claude/skills/
    ├── ai-multimodal/
    │   └── gemini-vision/
    ├── databases/
    │   └── postgresql-psql/
    └── devops/
        └── cloudflare-dns/

Customizations in any skill are detected and preserved automatically.

Development

See Development Guide for:

  • Project structure (modular domain-driven architecture)
  • Build & compilation (bun run build, bun run compile)
  • Testing & type checking
  • Code standards & linting

Architecture Highlights:

  • Modular design: 122 focused modules (target: <100 lines each)
  • Facade pattern: Each domain exposes public API via facade
  • Phase handlers: Complex commands use orchestrator + phase handlers
  • Self-documenting names: kebab-case file names describe purpose

Quick Start:

bun install
bun run dev new --kit engineer
bun test
# Optional: run expensive CLI integration tests explicitly
bun run test:integration

E2E Tests

Playwright E2E tests cover the gck migrate dashboard (3 scenarios). Tests run against the local dev server and use API mocking — no real filesystem state is modified.

Prerequisites: Node 18+ or Bun 1.0+, Chromium (installed automatically).

# One-time browser setup (if not already installed)
./node_modules/.bin/playwright install chromium

# Run all E2E specs
bun run test:e2e

# Interactive UI mode (watch + trace viewer)
bun run test:e2e:ui

Note: The dev server starts automatically via bun run dashboard:dev. CI wiring is a separate follow-up (local-only for now).

FAQ

Q: Do I need GitHub CLI? A: Yes, GitHub CLI is required. ClaudeKit uses it exclusively for authentication with private repositories.

Q: How do I authenticate? A: Run gh auth login, select 'Login with a web browser', complete OAuth in browser. Do NOT use 'Paste an authentication token'.

Q: "Access denied" error? A: Accept GitHub repo invitation, re-run gh auth login with web browser login, wait 2-5min for permissions.

Q: "GitHub CLI not authenticated" error? A: Run gh auth login and select 'Login with a web browser' (NOT 'Paste token'). PAT authentication is no longer supported.

Q: Is my token secure? A: Yes. GitHub CLI manages tokens securely via OAuth, stored encrypted in OS keychain.

License

MIT