One script. Zero duplicates. Every tool your freelance stack needs.
A complete, automated setup for full-stack freelancers using Claude Code, Cursor, Aider, Windsurf, claude.ai Web, and Claude Desktop. Installs every tool, framework, and AI skill you need β skips what you already have.
Quick Start Β· Skills Β· Web & Desktop Β· Doctor Β· Scaffold Β· Token Tracking Β· Universal AI Β· Contributing
Setting up a new Mac for full-stack freelance work takes hours β installing Flutter, Node, Docker, configuring VS Code, setting up Claude skills. This repo automates everything in one script:
- β Smart install β checks before installing, never duplicates
- β Safe to re-run β run anytime to add new tools or catch missing ones
- β 42 Claude AI skills β auto-trigger the right expert for every task
- β Interactive skill selector β choose exactly what you need by category
- β
Health check (
./doctor.sh) β verify your entire dev environment likeflutter doctor - β Project scaffolding β generate Next.js, NestJS, Expo, Flutter, or fullstack monorepo projects
- β Token & cost tracking β see input/output tokens and cost per request
- β Token optimizer β built-in model selection, context management, and prompt efficiency to maximise every token
- β Universal AI support β works with Claude Code, Cursor, Aider, and Windsurf
- β Web & Desktop support β export skills as system prompts for claude.ai Projects and Claude Desktop
- β Linux support β Ubuntu, Debian, and derivatives
- β
Auto-update β pull latest skills and configs with
./update.sh - β .env generator β generate environment templates for any stack
- β GitHub Actions templates β CI/CD pipelines ready to copy
- β VS Code fully configured β settings, keybindings, 21 extensions
- β Freelancer-ready β Upwork proposals, client replies, project scoping built in
Requires: macOS, Linux, or Windows Β· Internet connection Β· ~15 min
# Clone the repo
git clone https://github.com/thesaifalitai/claude-setup.git
cd claude-setup
# macOS (Apple Silicon & Intel)
chmod +x setup.sh && ./setup.sh
# Linux (Ubuntu/Debian)
chmod +x setup-linux.sh && ./setup-linux.sh
# Windows (PowerShell β Run as Administrator)
powershell -ExecutionPolicy Bypass -File .\setup-windows.ps1
# Verify your environment
./doctor.shThat's it. The script handles everything else β including skipping tools you already have.
# All commands use -ExecutionPolicy Bypass to avoid signing errors
powershell -ExecutionPolicy Bypass -File .\setup-windows.ps1 # Interactive β choose what to install
powershell -ExecutionPolicy Bypass -File .\setup-windows.ps1 -All # Install everything (no prompts)
powershell -ExecutionPolicy Bypass -File .\setup-windows.ps1 -SkillsOnly # Only install Claude skills (no dev tools)
powershell -ExecutionPolicy Bypass -File .\setup-windows.ps1 -Only node,git # Install specific tools only
powershell -ExecutionPolicy Bypass -File .\setup-windows.ps1 -List # Show available tools & install statusNote: Windows script uses
winget(built-in on Windows 10 1709+). For tools not in winget, it falls back tonpmorChocolatey. The-ExecutionPolicy Bypassflag is needed because unsigned scripts are blocked by default on Windows.
curl -fsSL https://raw.githubusercontent.com/thesaifalitai/claude-setup/main/setup.sh | bash
β οΈ The one-liner won't copy skills locally. Use thegit clonemethod to get all 43 skills installed.
The setup script automatically installs two hooks + optimal settings:
| Hook | What it does |
|---|---|
compact-reminder.sh |
Reminds you to /compact at turns 20, 40, 60+ β prevents context overflow |
auto-skill.sh |
Detects the file type you're editing and suggests the right skill automatically |
Settings applied:
| Setting | Value | Benefit |
|---|---|---|
CLAUDE_AUTOCOMPACT_PCT_OVERRIDE |
50 |
Auto-compacts at 50% context usage |
CLAUDE_CODE_SUBAGENT_MODEL |
haiku |
Subagents use Haiku (3Γ cheaper) |
MAX_THINKING_TOKENS |
10000 |
Caps internal reasoning (faster + focused) |
To install hooks separately:
./install_hooks.shEach item is checked before installing β already installed = skipped with β in the output.
| # | Tool | Version | Purpose |
|---|---|---|---|
| 1 | Xcode CLI Tools | latest | macOS build tools |
| 2 | Homebrew | latest | macOS package manager |
| 3 | Git | latest | Version control |
| 4 | GitHub CLI (gh) |
latest | GitHub from terminal |
| 5 | NVM | latest | Node version manager |
| 6 | Node.js | 20 LTS | JavaScript runtime |
| 7 | Claude CLI | latest | AI coding assistant |
| 8 | TypeScript | latest | Type-safe JavaScript |
| 9 | tsx | latest | Run TS files directly |
| 10 | NestJS CLI | latest | Backend scaffolding |
| 11 | EAS CLI | latest | Expo app builds |
| 12 | Expo CLI | latest | React Native tooling |
| 13 | Flutter + Dart | latest stable | Cross-platform mobile |
| 14 | Docker Desktop | latest | Containerization |
| 15 | PostgreSQL 16 | 16.x | Relational database |
| 16 | Redis | 7.x | Cache & queue |
| 17 | Nginx | latest | Web server / proxy |
| 18 | AWS CLI | latest | Cloud infrastructure |
| 19 | VS Code | latest | Code editor |
Claude AI automatically uses the right skill based on what you type. No manual activation needed.
react-native-expo β React Native + Expo (click to expand)
Triggers on: React Native, Expo, Expo Router, EAS Build, iOS/Android app, mobile dev, Reanimated, FlashList
What it knows:
- Expo SDK 50+ and bare CLI workflows
- Expo Router (file-based navigation)
- EAS Build + EAS Submit + OTA updates
- Performance: FlashList, Reanimated 3, Skia
- State: Zustand + React Query
- GitHub Actions β EAS Build CI/CD
- Sentry crash reporting
# Example triggers
"Build a React Native login screen with biometric auth"
"Set up Expo Router with tabs and auth flow"
"Write an EAS Build GitHub Actions workflow"flutter-dev β Flutter + Dart (click to expand)
Triggers on: Flutter, Dart, Riverpod, GoRouter, Freezed, Material 3, Flutter test
What it knows:
- Clean architecture (data / domain / presentation)
- Riverpod 2.x state management with code generation
- GoRouter file-based navigation
- Freezed data models + JSON serialization
- Dio HTTP client with interceptors & retry
- Fastlane + Codemagic CI/CD
# Example triggers
"Create a Flutter app with Riverpod authentication"
"Set up GoRouter with auth guards in Flutter"
"Write a Dio interceptor with refresh token logic"flutter-expert / react-native-expert β Community skills from Jeffallan
Additional expert-level mobile skills sourced from Jeffallan/claude-skills β focused on architecture patterns and production best practices.
nodejs-backend β Node.js / NestJS / Express
Triggers on: Node.js, NestJS, Express, REST API, GraphQL, JWT, Prisma, Redis, BullMQ
What it knows:
- NestJS module architecture (controllers, services, guards)
- Prisma ORM with PostgreSQL/MySQL schemas
- JWT authentication with refresh token rotation
- Redis caching (ioredis)
- BullMQ job queues
- Global exception filters + interceptors
- Rate limiting + Helmet security
"Build a NestJS CRUD API for products with JWT auth"
"Create a Prisma schema for a multi-tenant SaaS"
"Set up BullMQ email queue with retry logic"nestjs-expert β NestJS deep expertise (community)
Enterprise-grade NestJS patterns including dependency injection, custom decorators, microservices, and testing strategies.
django-expert β Django + DRF
Triggers on: Django, DRF, Python web app, Django REST framework, Django ORM
Production Django apps with Django REST Framework, Celery, PostgreSQL, and deployment patterns.
fastapi-expert β FastAPI async Python
Triggers on: FastAPI, async Python API, Pydantic, Python REST API
Async Python APIs with FastAPI, Pydantic V2, SQLAlchemy async, and Alembic migrations.
laravel-specialist β Laravel PHP
Triggers on: Laravel, PHP, Eloquent, Artisan, Blade
Modern Laravel with Eloquent ORM, Livewire, Sanctum auth, and Forge deployment.
nextjs-frontend β Next.js + React + Vue
Triggers on: Next.js, App Router, React, Vue, SSR, SSG, Vercel, shadcn/ui, React Query
What it knows:
- Next.js 14 App Router (Server vs Client components)
- React Hook Form + Zod validation
- TanStack Query (React Query) data fetching
- Next.js Metadata API (SEO)
- Middleware (auth protection)
- Core Web Vitals optimization
"Build a Next.js dashboard with server components and Prisma"
"Create a login form with React Hook Form and Zod"
"Optimize my Next.js app for Core Web Vitals"nextjs-developer / vue-expert β Community skills
Deep-dive Next.js Server Actions, Server Components, and Vue 3 Composition API with Pinia.
typescript-pro β Advanced TypeScript
Triggers on: TypeScript generics, utility types, type guards, conditional types, mapped types
Expert TypeScript: advanced generics, template literal types, discriminated unions, branded types.
uiux-design β Tailwind + Component Design Systems
Triggers on: UI design, Tailwind CSS, shadcn/ui, dark mode, accessibility, responsive, animations
What it knows:
- Design tokens with CSS variables (light/dark)
- CVA (class-variance-authority) component variants
- Framer Motion page transitions + staggered lists
- WCAG 2.1 AA accessibility patterns
- Skeleton loaders + empty states
- Mobile-first responsive layouts
"Design a responsive dashboard sidebar with dark mode"
"Create an accessible modal with focus management"
"Build a design system with Tailwind and shadcn/ui"ui-ux-pro-max β Design Intelligence (67 styles Β· 96 palettes Β· 57 fonts Β· 100 reasoning rules)
Triggers on: design, build, create, implement, review, fix, improve UI/UX for websites, landing pages, dashboards, SaaS, mobile apps, e-commerce, portfolios
What it knows:
- 67 UI styles β glassmorphism, claymorphism, brutalism, bento grid, neumorphism and more
- 96 industry-specific color palettes with hex values and mood context
- 57 Google Font pairings with CSS imports
- 100 reasoning rules mapping product type β optimal style/color/typography
- 99 UX best practices (accessibility, animations, touch targets)
- 25 chart types with library recommendations
- Stack-specific guidelines: React, Next.js, Vue, Flutter, React Native, SwiftUI, shadcn/ui, Jetpack Compose
How it works: Runs a Python BM25 search engine over local CSV databases to generate a complete design system β then uses that as a blueprint for code generation.
# Claude automatically runs:
python3 ~/.claude/skills/ui-ux-pro-max/scripts/search.py "beauty spa wellness" --design-system -p "Serenity Spa"
# Then generates pixel-perfect code using the design system outputRequires Python 3 β install with
brew install python3(macOS) orapt install python3(Linux)
devops-cicd β Docker / GitHub Actions / AWS / Nginx
Triggers on: Docker, Nginx, GitHub Actions, AWS, CI/CD, deployment, containers
What it knows:
- Multi-stage Dockerfiles (Node, Python)
- Docker Compose (dev + prod)
- GitHub Actions: test β build β push β deploy
- Nginx: SSL, reverse proxy, gzip, security headers
- AWS Lambda (Serverless Framework)
- Blue/green zero-downtime deployments
- Winston structured logging + Sentry
"Write a multi-stage Dockerfile for my NestJS app"
"Create a GitHub Actions workflow that deploys to EC2"
"Configure Nginx as a reverse proxy with SSL"kubernetes-specialist / cloud-architect β K8s + Multi-cloud
Kubernetes deployments, Helm charts, Horizontal Pod Autoscaling, AWS/GCP/Azure architecture patterns.
devops-engineer β Community DevOps skill
Platform engineering, advanced CI/CD pipelines, and production incident response patterns.
fullstack-architecture β System Design + Monorepo
Triggers on: Architecture, monorepo, Turborepo, system design, DB schema, multi-tenant SaaS
What it knows:
- Turborepo monorepo (web + api + mobile in one repo)
- Multi-tenant SaaS schema patterns
- REST API design standards (pagination, errors)
- JWT + refresh token rotation architecture
- S3 presigned URL upload flow
- Stripe payment integration pattern
- WebSocket + SSE real-time patterns
- Redis caching strategy (TTLs, invalidation)
graphql-architect / microservices-architect / api-designer
GraphQL schema design + federation, microservices with service mesh, REST API OpenAPI standards.
upwork-freelancer β Proposals + Client Communication
Triggers on: Upwork proposal, bid, client message, project scope, quote, client reply, freelance
Templates included:
- 3 winning proposal templates (React Native, Node.js, Full Stack)
- Discovery call message
- Project kickoff message
- Scope change request
- Asking for a review
- Handling vague requirements
- Rate negotiation scripts
- Project scoping document template
- Red flags checklist
"Write an Upwork proposal for this React Native job: [paste JD]"
"Reply to this client who wants to change scope mid-project"
"Write a project kickoff message for my new client"token-tracker β Token Usage & Cost Monitoring
Triggers on: token usage, API cost, token count, input/output tokens, billing, cost tracking, cost estimate
What it knows:
- Per-request token estimation (input + output + context %)
- Cost calculation for Claude Opus/Sonnet/Haiku with cache pricing
- Cross-provider cost comparison (GPT-4o, Gemini 2.0, DeepSeek)
- Session summaries with cumulative stats
- Context window usage tracker with /compact warnings
- Daily / monthly budget estimator by use case
"Enable token tracking"
"How much did this request cost?"
"Compare costs across Claude models"
"Show me my session token usage summary"
"Estimate my monthly cost if I use Claude 2 hours/day"token-optimizer β Active Token Saving & Context Efficiency
Triggers on: save tokens, reduce cost, context window full, which model should I use, Haiku vs Sonnet vs Opus, when to compact, prompt is too long, Claude is getting slow
What it knows:
- Model selection guide β exact task β model mapping (Haiku / Sonnet / Opus)
- When and how to use
/compact,/clear, and subagents - Prompt engineering templates that eliminate wasted tokens
- Context inclusion strategy (what to include vs exclude)
- Subagent strategy to keep main context clean
- Prompt caching patterns (90% input cost reduction)
- Session efficiency checklist
- Emergency context-full recovery procedure
"Which model should I use for code generation?"
"How do I save tokens when working with large files?"
"My context is almost full, what do I do?"
"Show me the proactive compaction schedule"
"How do I write prompts that use fewer tokens?"technical-writer β README Β· ADR Β· Runbook Β· OpenAPI Β· Changelog Β· Postmortem
Triggers on: README, API docs, ADR, runbook, changelog, postmortem, onboarding guide, technical blog, JSDoc, TSDoc
What it knows:
- README templates (Quick Start, API Reference, Configuration, Contributing)
- Architecture Decision Records (ADR) β captures why, not just what
- Runbooks designed for incident response (fast to scan under pressure)
- OpenAPI / Swagger documentation
- Changelogs in Keep a Changelog format
- Postmortem templates (timeline, root cause, action items)
- JSDoc / TSDoc annotation patterns
"Write a README for my Next.js SaaS starter"
"Create an ADR for switching from Mongoose to Prisma"
"Write a runbook for database connection failures"
"Generate a changelog for this release"
"Write a postmortem for last night's outage"cto-advisor β Tech Strategy Β· Team Scaling Β· Build vs Buy Β· DORA Metrics Β· Engineering Culture
Triggers on: technology strategy, engineering roadmap, build vs buy, technical debt, DORA metrics, team structure, engineering OKRs, hiring engineers, vendor evaluation, first 90 days as CTO
What it knows:
- Tech Radar (Adopt / Trial / Assess / Hold)
- Build vs Buy decision framework
- OKR templates for engineering teams
- Technical debt classification matrix (2Γ2: impact vs effort)
- DORA metrics benchmarks (elite, high, medium, low)
- Team Topologies (stream-aligned, platform, enabling, complicated-subsystem)
- Engineering hiring scorecard
- First 90 days as CTO / tech lead playbook
- Blameless postmortem culture
- Vendor / platform evaluation template
"Should we build our own auth or use Auth0?"
"Help me write engineering OKRs for next quarter"
"How should I structure my engineering team as we go from 5 to 20 engineers?"
"What are the DORA metrics and how do I measure them?"
"Write a vendor evaluation for choosing between Supabase and PlanetScale"project-manager β Agile Β· Sprint Planning Β· Stakeholder Reports Β· Scope Management Β· Release Planning
Triggers on: sprint planning, backlog grooming, user stories, retrospective, stakeholder update, scope creep, milestone planning, status report, release checklist, project kick-off
What it knows:
- Agile frameworks comparison (Scrum, Kanban, Shape Up, Dual-track)
- User story format (As a / I want / So that + Acceptance Criteria + DoD)
- Sprint planning template (capacity, velocity, backlog)
- Sprint retrospective (Start / Stop / Continue)
- Project kick-off document (RACI, milestones, risks, communication plan)
- Weekly status report template
- Scope change management (impact assessment + client templates)
- Release planning checklist (engineering + product + process)
"Write a project kick-off document for our payments feature"
"Help me structure this sprint retrospective"
"A client wants to add a feature mid-sprint β how do I handle it?"
"Write a weekly status report for our mobile app project"
"Create a release checklist for our v2.0 launch"task-estimator β Story Points Β· Epic Breakdown Β· Sprint Capacity Β· Timeline Forecasting Β· Spike Templates
Triggers on: estimation, story points, task breakdown, how long, timeline, sprint capacity, epic decomposition, T-shirt sizing, three-point estimation, Monte Carlo, spike
What it knows:
- Three-point estimation (Optimistic / Most Likely / Pessimistic)
- Epic decomposition: vertical slicing (not horizontal)
- Hidden work checklist (error states, tests, migrations, reviews)
- Story point reference card (1β13, with wall-clock guidance)
- Sprint capacity formula (70% effective capacity rule)
- Timeline forecasting: velocity-based + Monte Carlo
- Estimation bias detection (optimism bias, planning fallacy, novelty tax)
- Stakeholder communication templates for estimates and bad news
- Spike templates for de-risking unknowns
"Break down this 'user authentication' epic into stories and estimate it"
"How long will it take to build a Stripe subscription billing system?"
"We have 4 engineers for a 2-week sprint β what's our capacity?"
"A stakeholder is asking when the feature will be done β help me answer"
"This is too uncertain to estimate β write a spike for it"debugging-wizard / code-reviewer / test-master / secure-code-guardian
| Skill | Purpose |
|---|---|
debugging-wizard |
Systematic debugging across all languages |
code-reviewer |
Thorough code reviews with actionable feedback |
test-master |
Unit, integration, E2E, and performance testing |
secure-code-guardian |
Security vulnerabilities, OWASP, secure patterns |
feature-forge |
Requirements gathering β technical spec |
The script automatically configures VS Code with:
| Category | Extensions |
|---|---|
| Core | Prettier, ESLint, GitLens, Error Lens |
| React Native | React Native Tools, Expo Tools |
| Flutter | Dart, Flutter |
| Backend | Path IntelliSense, DotENV, Thunder Client |
| Database | Prisma |
| DevOps | Docker |
| UI | Tailwind CSS IntelliSense |
| Themes | Material Theme, Material Icons |
| Testing | Jest |
| Productivity | Todo Tree, Better Comments |
- Format on save (Prettier)
- ESLint auto-fix on save
- TypeScript strict mode
- JetBrains Mono font with ligatures
- Tab size: 2 spaces
- Dark mode ready
- Language-specific formatters (Dart uses its own)
claude-setup/
βββ π README.md β You are here
βββ π CONTRIBUTING.md β How to add skills or report issues
βββ π CHANGELOG.md β Version history
βββ π RECOMMENDED_SKILLS.md β Voted skills roadmap
βββ π LICENSE β MIT License
βββ π .gitignore
βββ π CLAUDE.md β Global Claude coding rules
βββ π .cursorrules β Cursor AI config
βββ π .aider.conf.yml β Aider AI config
βββ π .windsurfrules β Windsurf AI config
βββ βοΈ setup.sh β Main install script (macOS)
βββ βοΈ setup-linux.sh β Linux install script (Ubuntu/Debian)
βββ βοΈ setup-windows.ps1 β Windows install script (Win 10/11, PowerShell)
βββ βοΈ install_skills.sh β Interactive skill selector
βββ βοΈ install_hooks.sh β Efficiency hooks installer
βββ βοΈ doctor.sh β Health check (like flutter doctor)
βββ βοΈ scaffold.sh β Project scaffolding generator
βββ βοΈ update.sh β Auto-update skills and configs
βββ βοΈ generate-env.sh β .env.example generator
βββ βοΈ universal-setup.sh β Multi-tool config copier
βββ βοΈ token-tracker.sh β Token & cost tracking utility
βββ βοΈ export-for-web.sh β Export skills for claude.ai Web & Desktop
βββ π WEB_DESKTOP_GUIDE.md β Guide for Web & Desktop skill setup
β
βββ πͺ hooks/ β Auto-run hooks for efficiency
β βββ compact-reminder.sh β /compact reminder at turns 20/40/60+
β βββ auto-skill.sh β Suggest the right skill per file type
β
βββ βοΈ settings/
β βββ efficiency.json β Optimal settings template
β
βββ π§ skills/ β 42 Claude Code skills
β βββ react-native-expo/
β βββ flutter-dev/
β βββ nodejs-backend/
β βββ nextjs-frontend/
β βββ uiux-design/
β βββ ui-ux-pro-max/ β Design intelligence (67 styles, 96 palettes, BM25 search)
β βββ devops-cicd/
β βββ upwork-freelancer/
β βββ fullstack-architecture/
β βββ supabase-expert/ β NEW: Supabase (RLS, Edge Functions, Auth)
β βββ stripe-expert/ β NEW: Payments (Checkout, Subscriptions, Webhooks)
β βββ prisma-expert/ β NEW: ORM (Schema, Queries, Migrations)
β βββ ai-integration/ β NEW: LLM integration (RAG, Streaming, Claude API)
β βββ auth-patterns/ β NEW: Auth (NextAuth, JWT, RBAC, OAuth)
β βββ token-tracker/ β Token & cost tracking skill
β βββ token-optimizer/ β Model selection, context mgmt, prompt efficiency
β βββ technical-writer/ β README, ADR, runbook, postmortem, changelog
β βββ cto-advisor/ β Tech strategy, team scaling, DORA metrics
β βββ project-manager/ β Agile, sprint planning, stakeholder reports
β βββ task-estimator/ β Story points, epic breakdown, timeline forecasting
β βββ ... (22 more from community)
β
βββ π web-prompts/ β Pre-exported skill prompts (run export-for-web.sh)
β
βββ π templates/
β βββ github-actions/ β CI/CD pipeline templates
β β βββ ci-nextjs.yml
β β βββ ci-nestjs.yml
β β βββ deploy-vercel.yml
β β βββ docker-build.yml
β βββ env-examples/ β .env templates
β
βββ π» vscode/
β βββ settings.json β VS Code settings
β βββ keybindings.json β Keyboard shortcuts
β βββ extensions.txt β Extension list
β
βββ π .github/
βββ workflows/
β βββ ci.yml β Validates skills on push
βββ ISSUE_TEMPLATE/
βββ bug_report.md
βββ feature_request.md
# Pull latest skills and tools list
cd claude-setup
git pull origin main
# Re-run setup β only installs what's new/missing
./setup.shWe welcome new skills, tool updates, and improvements!
See CONTRIBUTING.md for full guidelines.
Quick contribution:
- Fork this repo
- Add your skill to
skills/your-skill-name/SKILL.md - Submit a Pull Request
Skill ideas welcome:
python-proimprovementsstripe-integrationskillsupabaseskillprisma-expertskill- Windows/Linux support for
setup.sh
Frontend: React Native Β· Flutter Β· Next.js Β· React.js Β· Vue.js Β· TypeScript Β· Tailwind CSS
Backend: Node.js Β· Python Β· Django Β· Laravel Β· NestJS Β· Express.js Β· FastAPI Β· GraphQL
Databases: MySQL Β· MongoDB Β· PostgreSQL Β· Pinecone
Cloud: AWS (Lambda, EC2, S3) Β· Docker Β· Kubernetes Β· Nginx Β· Firebase
Freelance: Upwork Β· LinkedIn Β· Fiverr Β· Proposals Β· Cover Letters Β· Client Communication
Don't need all 42 skills? Pick only what you want:
# Interactive menu β choose by category
./install_skills.sh
# Install everything (no prompts)
./install_skills.sh --all
# Install by category
./install_skills.sh --category mobile # React Native + Flutter
./install_skills.sh --category backend # Node.js, NestJS, Django, Supabase, Stripe, Prisma, Auth
./install_skills.sh --category frontend # Next.js, Vue, TypeScript
./install_skills.sh --category devops # Docker, K8s, AWS, CI/CD
./install_skills.sh --category quality # Testing, reviews, security
./install_skills.sh --category freelance # Upwork proposals + cover letters
# Install a single skill
./install_skills.sh --skill proposal-writer # Just the proposal writer
./install_skills.sh --skill supabase-expert
# Search for skills by keyword
./install_skills.sh --search payment # Find payment-related skills
./install_skills.sh --search auth # Find auth-related skills
# Remove a skill
./install_skills.sh --remove flutter-expert
# See what's available
./install_skills.sh --listCategories available: mobile, backend, frontend, uiux, devops, architecture, quality, languages, freelance, utilities, management
Don't want the full repo? Grab only the skills you need with a single command.
Replace SKILL_NAME with any skill from the skills list:
macOS / Linux / WSL / Git Bash:
SKILL_NAME="proposal-writer" # change to any skill name
mkdir -p ~/.claude/skills/$SKILL_NAME
curl -fsSL "https://raw.githubusercontent.com/thesaifalitai/claude-setup/main/skills/$SKILL_NAME/SKILL.md" \
-o ~/.claude/skills/$SKILL_NAME/SKILL.mdWindows (PowerShell):
$skill = "proposal-writer" # change to any skill name
New-Item -ItemType Directory -Force -Path "$env:USERPROFILE\.claude\skills\$skill"
Invoke-WebRequest -Uri "https://raw.githubusercontent.com/thesaifalitai/claude-setup/main/skills/$skill/SKILL.md" `
-OutFile "$env:USERPROFILE\.claude\skills\$skill\SKILL.md"macOS / Linux / WSL:
for skill in proposal-writer upwork-freelancer supabase-expert; do
mkdir -p ~/.claude/skills/$skill
curl -fsSL "https://raw.githubusercontent.com/thesaifalitai/claude-setup/main/skills/$skill/SKILL.md" \
-o ~/.claude/skills/$skill/SKILL.md
echo "installed: $skill"
doneWindows (PowerShell):
foreach ($skill in @("proposal-writer", "upwork-freelancer", "supabase-expert")) {
New-Item -ItemType Directory -Force -Path "$env:USERPROFILE\.claude\skills\$skill"
Invoke-WebRequest -Uri "https://raw.githubusercontent.com/thesaifalitai/claude-setup/main/skills/$skill/SKILL.md" `
-OutFile "$env:USERPROFILE\.claude\skills\$skill\SKILL.md"
Write-Host "installed: $skill"
}- Browse skills/ folder on GitHub
- Open the skill folder you want
- Download
SKILL.md - Place it at
~/.claude/skills/<skill-name>/SKILL.md
No scripts, no dependencies β each skill is a single
SKILL.mdfile that works independently.
| Platform | How |
|---|---|
| Claude Code (CLI) | Auto-detected from ~/.claude/skills/ |
| VS Code extension | Auto-detected from ~/.claude/skills/ |
| Claude.ai Web | Copy skill content into Settings > Custom Instructions |
| Claude Desktop App | Copy skill content into Settings > Custom Instructions |
Verify your entire development environment with a single command β like flutter doctor for your whole stack:
./doctor.shβββ System ββββββββββββββββββββββββββββ
[β] macOS 14.2 (arm64)
[β] Disk space: 85GB available
βββ Core Tools ββββββββββββββββββββββββ
[β] Git 2.43.0
[β] GitHub CLI 2.42.0 β authenticated
βββ Node.js Ecosystem βββββββββββββββββ
[β] Node.js v20.11.0
[β] npm 10.2.4
βββ Claude Code βββββββββββββββββββββββ
[β] Claude CLI 1.0.3
[β] Claude skills: 36 installed
βββ Infrastructure ββββββββββββββββββββ
[β] Docker 24.0.7 β daemon running
[β] PostgreSQL 16.1 β accepting connections
[β] Redis 7.2.3 β responding
Summary: 22 passed 3 warnings 0 failed
Checks: OS, disk space, Git, GitHub CLI, Node.js, npm, NVM, Claude CLI, skills, Flutter, Docker, PostgreSQL, Redis, Nginx, AWS CLI, VS Code, and all AI tool configs.
Generate a fully configured project in seconds β all configs, env files, and AI tools pre-wired:
# Interactive menu
./scaffold.sh
# Direct command
./scaffold.sh nextjs my-saas-app # Next.js + Tailwind + Prisma + NextAuth
./scaffold.sh nestjs my-api # NestJS + Prisma + JWT + Swagger
./scaffold.sh expo my-mobile-app # React Native + Expo Router + NativeWind
./scaffold.sh flutter my_app # Flutter + Riverpod + GoRouter
./scaffold.sh fullstack my-platform # Monorepo: Next.js + NestJS + shared typesEvery scaffolded project includes:
- All AI tool configs (CLAUDE.md, .cursorrules, .aider.conf.yml, .windsurfrules)
.env.examplewith all required variables- Git initialized with first commit
- Docker Compose for local services (fullstack)
Keep your skills and configs up to date:
# Update everything (repo + skills + configs)
./update.sh
# Update skills only
./update.sh --skills
# Update CLAUDE.md only
./update.sh --configs
# Check installed version
./update.sh --versionGenerate .env.example templates for any stack:
# Interactive menu
./generate-env.sh
# Direct command
./generate-env.sh nextjs ./my-project # Next.js env template
./generate-env.sh nestjs ./my-api # NestJS env template
./generate-env.sh supabase . # Supabase env template
./generate-env.sh fullstack . # Fullstack monorepo
./generate-env.sh expo . # React Native / ExpoFull Linux support via setup-linux.sh β works on Ubuntu 22.04+, Debian 12+, and derivatives:
# Linux install
chmod +x setup-linux.sh && ./setup-linux.sh
# Then verify
./doctor.shInstalls the same 19 tools using apt instead of Homebrew, with proper systemd service management.
Ready-to-copy CI/CD pipelines in templates/github-actions/:
| Template | File | What It Does |
|---|---|---|
| Next.js CI | ci-nextjs.yml |
Lint, type check, test, build with PostgreSQL |
| NestJS CI | ci-nestjs.yml |
Lint, test, e2e test with PostgreSQL + Redis |
| Vercel Deploy | deploy-vercel.yml |
Auto-deploy to Vercel on push to main |
| Docker Build | docker-build.yml |
Build and push to GitHub Container Registry |
# Copy a template to your project
cp templates/github-actions/ci-nextjs.yml your-project/.github/workflows/ci.ymlTrack how many tokens each AI request uses and what it costs β and actively reduce them.
Two complementary skills:
| Skill | Purpose |
|---|---|
token-tracker |
Measure β per-request cost, session totals, budget estimates |
token-optimizer |
Reduce β model selection, context management, prompt efficiency |
./install_skills.sh --skill token-tracker
./install_skills.sh --skill token-optimizer# Tracking β in any Claude Code session:
"Enable token tracking"
"How much did this request cost?"
"Show me a session summary"
"Estimate my monthly cost for heavy use"
# Optimization:
"Which model should I use for this task?"
"My context is almost full β what do I do?"
"How do I write prompts that use fewer tokens?"
"Show me the model selection guide"Claude will append token usage after every response and proactively suggest /compact when context gets long.
The included CLAUDE.md encodes token-saving behaviors that are always active β no skill needed:
- No preamble or filler phrases
- Code over prose (show the diff, not the explanation)
- Suggest
/compactafter major phases,/clearbetween tasks - Use subagents for file exploration (keeps main context clean)
- Read files with line ranges, not entire files
# Load the tracker in your terminal
source token-tracker.sh
# Start tracking
tt_enable
# After each AI interaction, log character counts
tt_log 2000 5000 "code review request"
# See session totals
tt_summary
# Compare costs across providers
tt_compare
# Set your model for accurate pricing
tt_set_model opus # or: sonnet, haiku
# View past sessions
tt_historyCost comparison across providers:
π° Cost Comparison (~1,000 in / ~2,000 out tokens)
Claude Opus 4: $0.165000
Claude Sonnet 4: $0.033000
Claude Haiku 3.5: $0.008800
GPT-4o: $0.022500
Gemini 1.5 Pro: $0.011250
DeepSeek V3: $0.002470
This repo works with all major AI coding tools, not just Claude Code. One set of coding standards, multiple tools.
# Copy configs to your project for ALL tools at once
./universal-setup.sh /path/to/your/project
# Or manually copy what you need:
cp CLAUDE.md /your/project/ # Claude Code
cp .cursorrules /your/project/ # Cursor
cp .aider.conf.yml /your/project/ # Aider
cp .windsurfrules /your/project/ # Windsurf| AI Tool | Config File | Status |
|---|---|---|
| Claude Code (priority) | CLAUDE.md + ~/.claude/skills/ |
Full support (skills + config) |
| claude.ai Web | Project Instructions | Full support (export-for-web.sh) |
| Claude Desktop | Project Instructions | Full support (export-for-web.sh) |
| Cursor | .cursorrules |
Full support (coding standards) |
| Aider | .aider.conf.yml |
Full support (model + settings) |
| Windsurf (Codeium) | .windsurfrules |
Full support (coding standards) |
Claude Code is our primary focus β it gets the full skill system with auto-triggering. Other tools get the shared coding standards. claude.ai Web and Desktop get the full skill knowledge via Project Instructions (see below).
Skills work on claude.ai and Claude Desktop via Project Instructions. Export any skill (or a combination) as a ready-to-paste system prompt:
# Interactive picker β select skills by number
./export-for-web.sh
# Export a single skill (auto-copies to clipboard)
./export-for-web.sh --skill flutter-dev
# Export multiple skills into one combined prompt
./export-for-web.sh --skill nextjs-developer --skill uiux-design --skill secure-code-guardian
# Export all skills at once
./export-for-web.sh --all
# See what's available
./export-for-web.sh --listThen paste into a Project:
- claude.ai β Projects β β Settings β Project Instructions β paste β Save
- Claude Desktop β Projects β Edit β Custom Instructions β paste β Save
Every conversation inside that Project will have the skill expertise active β no manual invocation needed.
Recommended Projects:
| Project | Skills |
|---|---|
| Mobile Dev | react-native-expo + flutter-dev |
| Full-Stack Web | nextjs-developer + nextjs-frontend + nodejs-backend |
| UI/UX | uiux-design + ui-ux-pro-max |
| DevOps | devops-cicd + devops-engineer |
| Code Quality | code-reviewer + test-master + secure-code-guardian |
Note: The
ui-ux-pro-maxskill uses a Python BM25 search engine that only runs in Claude Code CLI. On Web/Desktop, Claude uses the embedded knowledge instead β design planning still works, just without the live database search.
See WEB_DESKTOP_GUIDE.md for the full setup walkthrough.
Q: Is it safe to re-run setup.sh?
A: Yes. Every install is guarded by a check. Already-installed items print β already installed and are skipped.
Q: Does this work on Intel Macs?
A: Yes. The script detects arm64 vs x86_64 and adjusts Homebrew paths accordingly.
Q: Can I use this on Linux / Windows?
A: Full setup script is macOS + Linux. Skills work on any OS (Windows, macOS, Linux) β just copy the skill folder to ~/.claude/skills/. Windows users can use PowerShell, WSL, or Git Bash (see install instructions above).
Q: What if a tool fails to install? A: The script prints a warning and continues. Failed installs are logged so you can fix them manually.
Q: Where are the skills stored?
A: ~/.claude/skills/ β global to your user, active in every Claude Code session.
Q: Can I use the skills on claude.ai or the Claude Desktop app?
A: Yes. Run ./export-for-web.sh --skill <name> to export a skill as a plain system prompt, then paste it into a Project's Instructions on claude.ai or Claude Desktop. See WEB_DESKTOP_GUIDE.md.
Q: How do I add my own skill?
A: Create skills/my-skill/SKILL.md, add a YAML frontmatter with name and description, then re-run setup.sh.
MIT β free to use, modify, and share. See LICENSE.
Skills sourced from:
- Jeffallan/claude-skills β 66 community skills
- affaan-m/everything-claude-code β Claude Code configs
- callstackincubator/agent-skills β React Native patterns
- nextlevelbuilder/ui-ux-pro-max-skill β Design intelligence skill (67 styles, 96 palettes, BM25 search engine)
- Anthropic Claude Code Docs β Official skill format
Made for freelancers who want to move fast without breaking things.
Works with: Claude Code Β· Cursor Β· Aider Β· Windsurf
β Star this repo if it saved you setup time!