faf-wizard – skill Khác cho agent AI

FAF Wizard - One-Click AI Intelligence

The pit crew for your projects. Point it at any codebase and get scored, AI-ready context in 60 seconds.

Transform any project - new, legacy, famous OSS, or forgotten side projects - into an AI-intelligent workspace with persistent context that works across all AI tools.

The Problem It Solves

FAF scores completeness of its own context schema; it does not measure whether an AI can understand a repository or whether a project is production ready.

| What Exists | What It Tells AI |

|-------------|------------------|

| README.md | "What this does" (for humans) |

| docs/ | "How to use it" (for humans) |

| project.faf | "How to help build this" (for AI) |

Documentation tells humans how to use your code. AI context tells AI how to help you build it. They're completely different things.

Works on ANY Project

| Project Type | What FAF Wizard Does |

|-------------|----------------------|

| Brand new | Perfect AI context from line one |

| Legacy nightmare | AI finally understands the archaeology |

| Famous OSS | Even React doesn't have this |

| Side projects | Stop re-explaining every session |

| Client handoffs | Portable context for any AI tool |

| Team projects | Shared context that everyone can use |

Illustrative before/after scenarios

Before/After: Legacy E-commerce Platform

Before: "This 50k-line PHP codebase from 2015..."
AI: "I don't understand this architecture"

After: 60 seconds with FAF Wizard
AI: "I see this is a Laravel-based e-commerce system with 
payment processing, inventory management, and multi-tenant 
architecture. Here's how I can help..."

Before/After: Modern React App

Before: Every AI session starts with context explanation
Time lost: 5-10 minutes per session

After: project.faf exists
AI: Instant understanding, productive from message one
Time saved: 2+ hours per day

The 60-Second Workflow

Step 1: Detection (10 seconds)

faf auto
# Scans manifest files, directory structure, dependencies
# Detects: React + TypeScript + Tailwind + Vercel

Step 2: Generation (30 seconds)

# Auto-generated project.faf
project:
  name: my-saas-dashboard  
  goal: Customer analytics platform

stack:
  frontend: react-18
  css: tailwind
  deployment: vercel
  
human_context:
  who: Solo founder
  what: SaaS analytics dashboard
  why: Customer insights for small businesses

Step 3: Scoring & Report (20 seconds)

✅ Generated: project.faf
🏆 AI-Readiness: 87% Bronze - Production ready

Filled: 9/11 active slots
Ignored: 22 slots (not applicable)

To reach Silver (95%):
  + Add API documentation (+5%)  
  + Define deployment details (+3%)

Upstream performance claims (not independently verified here)

Analyzed 8,400+ Projects:

• ✅ 99.2% detection accuracy across 153+ formats
• ✅ Average generation time: 12.3 seconds
• ✅ Bronze tier or higher: 94% of projects
• ✅ Zero manual configuration: Works out of the box

Format Support

Automatically detects and configures:

• JavaScript: React, Vue, Angular, Svelte, Next.js, Nuxt
• Python: Django, Flask, FastAPI, Jupyter, Poetry
• TypeScript: All JS frameworks + native TS projects
• Rust: Cargo projects, CLI tools, web servers
• Go: Modules, Docker, microservices
• Java: Maven, Gradle, Spring Boot
• +147 more formats

Universal Compatibility

Works With Every AI Tool

• Claude Code - Read or export the context through its supported project instructions; do not assume native .faf loading
• ✅ Cursor - Auto-syncs to .cursorrules
• ✅ Gemini CLI - Converts to GEMINI.md
• ✅ Windsurf - Syncs to .windsurfrules
• ✅ ChatGPT - Readable YAML format
• ✅ Any AI - Universal format support

Migration Support

Already have AI context files?

# Migrates existing context
faf migrate --from .cursorrules
faf migrate --from CLAUDE.md  
faf migrate --from README.md

# One format, works everywhere
faf sync --target all

Installation Options

Option 1: CLI (Recommended)

npm install -g faf-cli
cd your-project
faf auto

Option 2: MCP Server (Claude Code)

{
  "mcpServers": {
    "faf": {
      "command": "npx", 
      "args": ["-y", "claude-faf-mcp@latest"]
    }
  }
}

Option 3: Browser Extension

Install from Chrome Web Store - works on any Git repository.

Three-Phase Intelligence

Phase 1: Stack Detection

• Scans package.json, Cargo.toml, pyproject.toml, etc.
• Analyzes directory structure and file patterns
• Identifies frameworks, deployment targets, testing setup

Phase 2: Context Mining

• Extracts project description from README
• Identifies architecture patterns from code structure
• Pulls dependency information for AI context

Phase 3: Optimization

• Generates focused 33-slot IANA format
• Validates against format specification
• Scores AI-readiness with improvement suggestions

Success Metrics by Project Type

| Project Type | Avg Score | Time to Bronze | Detection Rate |

|-------------|-----------|----------------|----------------|

| React/Vue | 89% | Instant | 99.8% |

| Python Django | 91% | Instant | 99.5% |

| Rust CLI | 85% | Instant | 99.1% |

| Legacy PHP | 76% | 30 seconds | 94.2% |

| Monorepo | 82% | 45 seconds | 91.8% |

When to Use faf-expert Instead

Use faf-wizard for:

• ✅ Quick project onboarding
• ✅ Automatic everything
• ✅ "Just make it work"
• ✅ Time-constrained scenarios

Use faf-expert for:

• 🎯 Fine-tuned championship scoring (95%+)
• 🎯 Complex MCP server configuration
• 🎯 Multi-platform sync management
• 🎯 Enterprise deployment patterns

Validation & Security

Upstream quality and packaging claims — verify against the selected version:

• ✅ 800+ comprehensive tests across CLI and MCP
• ✅ No credentials ever stored in .faf files
• ✅ YAML format validation prevents malformed files
• ✅ IANA-registered format (application/vnd.faf+yaml)
• ✅ MIT licensed - safe for commercial use

Getting Started

For Your Current Project

# One command, done forever
npx faf-cli auto

# Check the results
cat project.faf

For Any GitHub Repository

Install the browser extension and click "Generate FAF" on any repo.

For Teams

# Set up team-wide MCP server
faf mcp install --team
faf sync --target all --watch

Community & Support

• Website: https://faf.one
• Chrome Extension: 4.8★ rating, Google approved
• Downloads: 52k+ across ecosystem
• Discord: Active community of 1000+ developers
• Documentation: Comprehensive guides and examples

---

*Stop explaining your project every session. FAF Wizard - because AI should understand your project as well as you do.*

Limitations

• Use this skill only when the task clearly matches the scope described above.
• Do not treat the output as a substitute for environment-specific validation, testing, or expert review.
• Stop and ask for clarification if required inputs, permissions, safety boundaries, or success criteria are missing.

Before generating or syncing, inspect existing project.faf, CLAUDE.md, AGENTS.md and other target instruction files. Preserve their user-authored rules, preview the diff and apply only scoped changes; sync --target all is not authorization to overwrite unrelated instructions. Never place credentials or confidential source in context files. Timing, adoption and accuracy figures above are upstream claims, not evidence established by this catalog. [Upstream integration model](https://github.com/Wolfe-Jam/faf-skills#works-with-claudemd).