MCPxelMCPxel
Co-Pilot
Updated a month ago

cc-guide

AAnsarUllahAnasZ360
0.0k
ansarullahanasz360/cc-guide
80
Agent Score

💡 Summary

A toolkit for autonomous AI-assisted development using Claude Code and Ralph loops.

🎯 Target Audience

Developers seeking to streamline their workflowFounders with innovative ideas but limited timeTeams looking to leverage AI without deep technical expertiseProject managers wanting to enhance productivityTech enthusiasts eager to explore AI capabilities

🤖 AI Roast: “Powerful, but the setup might scare off the impatient.”

Security AnalysisMedium Risk

Risk: Medium. Review: shell/CLI command execution; outbound network access (SSRF, data egress); filesystem read/write scope and path traversal; dependency pinning and supply-chain risk. Run with least privilege and audit before enabling in production.

Claude Code Guide

"Everything's coming up Milhouse!"

Skills and guides for autonomous AI-assisted development with Claude Code and Ralph loops.

What Is This?

You've seen the demos. People shipping entire products overnight. Loops running for hours. Features appearing while they slept.

And you've thought: "Why can't I do that?"

This repository is for you. The developer who feels one step behind. The founder with ideas but no time. The team where only one person "gets" AI.

Four skills that make the complicated stuff simple:

  • /prd — Create PRDs through conversation, not writing
  • /ralph-preflight — Validate everything before starting a Ralph loop
  • /setup-claude — Configure your repo optimally without reading 100 docs
  • /agent-browser — Browser automation for verifying UI actually works

Plus guides on how to actually use Claude Code productively.

Quick Start

Add Skills to Your Project

# Add all skills (recommended)
npx @ansarullahanas/cc-guide add-skill --all

# Or add individually
npx @ansarullahanas/cc-guide add-skill prd
npx @ansarullahanas/cc-guide add-skill ralph-preflight
npx @ansarullahanas/cc-guide add-skill setup-claude
npx @ansarullahanas/cc-guide add-skill agent-browser

Via skills.sh:

npx skills add AnsarUllahAnasZ360/cc-guide

Use the Skills

In Claude Code:

/prd                    # Create a PRD for a new feature
/ralph-preflight        # Validate config before starting Ralph loop
/setup-claude init      # Set up a new repository
/setup-claude audit     # Audit an existing repository
/agent-browser          # Browser automation help

The Big Idea

Traditional: You code, AI helps, you iterate forever.

This way: You define the work, AI executes, you review.

Traditional:          Ralph Loop:
Human → AI → Human    Human → PRD → AI Loop → Human Review
     ↑      ↓              ↓           ↓
     └──────┘          (autonomous)  (complete)

The quality of your PRD determines the quality of your output.

Why "Milhouse"?

In the world of AI-assisted development:

Ralph Wiggum is the AI — Confused, making mistakes, but never stopping. The AI is Ralph.

Milhouse is the human — Trying hard, a little confused, hoping this time it works.

This repository helps Milhouse win. Not by making you a prompt engineer, but by giving you skills that ask the right questions and make the right decisions.

Read the full story: What Is Milhouse?

What You Need

| Tool | Install | Why | |------|---------|-----| | Claude Code | claude.ai/claude-code | Core CLI | | GitHub CLI | brew install gh | Git operations | | tmux | brew install tmux | Persistent sessions | | Ralph TUI | npm install -g ralph-tui | Task orchestration | | Agent Browser | npm install -g agent-browser | Browser automation |

Guides

Start here:

| Guide | What You'll Learn | |-------|-------------------| | What Is Milhouse? | The philosophy behind this repo | | My Claude Workflow | How I actually use Claude Code daily | | Ralph Loops Explained | Running autonomous development loops | | Best Practices | Lessons learned the hard way |

Skills

PRD Skill (/prd)

Creates self-verifying PRDs through an interview process:

  1. Identifies what type of task you're doing
  2. Extracts everything you know via brain dump
  3. Asks clarifying questions
  4. Confirms understanding
  5. Generates PRD.md and prd.json

The output is a PRD that Ralph TUI can execute autonomously.

You don't write PRDs. You answer questions.

Ralph Pre-Flight Skill (/ralph-preflight)

Run this after /prd and before starting a Ralph loop:

  1. Detects global CLAUDE.md conflicts (can override local config)
  2. Checks for existing Ralph state from previous runs
  3. Validates config.toml and template paths
  4. Verifies prd.json structure and template variable mapping
  5. Offers branch vs worktree setup options
  6. Provides ready-to-copy tmux and Ralph launch commands

Catches configuration issues before they waste iterations.

Setup Claude Skill (/setup-claude)

Two modes:

Init — For new repositories:

  • Environment & tools check
  • CLI discovery & authentication
  • MCP configuration (context-window aware)
  • Skills, subagents, rules, hooks setup
  • CLAUDE.md generation
  • Ralph TUI configuration

Audit — For existing repositories:

  • Environment analysis
  • Gap analysis
  • Recommendations with interactive fixes
  • Context window optimization

Agent Browser Skill (/agent-browser)

Headless browser automation for AI agents:

  • Navigate, click, fill forms
  • Take screenshots and snapshots
  • E2E testing workflows
  • Web scraping

Essential for verifying that UI actually works, not just that Claude says it works.

Key Concepts

CLI First, MCP Last

CLIs don't consume context window. MCPs do.

Before enabling any MCP:

  • Check if a CLI exists (gh, vercel, supabase, stripe)
  • Install and authenticate the CLI
  • Only use MCP when there's no CLI alternative

Keep MCPs under 10. Under 5 is better.

tmux for Persistence

Ralph loops run for hours. tmux means:

  • Sessions persist through SSH disconnects
  • You can detach and come back later
  • Multiple loops can run in parallel
tmux new-session -s feature
ralph-tui run
# Ctrl+b, d to detach
# tmux attach -t feature to come back

Worktrees for Parallelism

Can't run two Ralph loops on the same codebase. Git worktrees fix this:

git worktree add ../project-feature-a feature-a

Each worktree can run its own tmux session with its own Ralph loop.

Repository Structure

cc-guide/
├── skills/
│   ├── prd/                  # PRD creation skill
│   │   ├── SKILL.md          # Entry point
│   │   ├── AGENTS.md         # Comprehensive reference
│   │   ├── interview/        # Interview process guides
│   │   └── categories/       # Task-type specific guidance
│   │
│   ├── ralph-preflight/      # Pre-flight check skill
│   │   └── SKILL.md          # Validation & launch commands
│   │
│   ├── setup-claude/         # Setup/audit skill
│   │   ├── SKILL.md          # Entry point
│   │   ├── workflows/        # Init & audit workflows
│   │   ├── components/       # Detailed component guides
│   │   ├── reference/        # Reference material
│   │   └── templates/        # Config templates
│   │
│   └── agent-browser/        # Browser automation skill
│       └── SKILL.md          # CLI reference
│
├── guides/
│   ├── what-is-milhouse.md        # Philosophy and purpose
│   ├── my-claude-workflow.md      # Daily workflow
│   ├── ralph-loops-explained.md   # Autonomous loops guide
│   └── best-practices.md          # Lessons learned
│
├── templates/
│   └── prompt.hbs            # Ralph TUI prompt template
│
└── bin/
    └── add-skill.js          # npx cc-guide add-skill

The Workflow

1. SETUP (once per project)
   /setup-claude init

2. PLAN (once per feature)
   /prd

3. VALIDATE (before each loop)
   /ralph-preflight

4. BUILD (autonomous)
   tmux new-session -s feature
   ralph-tui run --prd path/to/prd.json
   Ctrl+b, d

5. MONITOR (periodic)
   tmux attach -t feature

6. SHIP
   Review, merge, celebrate

Further Reading

Official:

Community:

License

MIT — See LICENSE

The goal isn't to remove humans from development. It's to elevate humans from typists to architects.

Everything's coming up Milhouse.

5-Dim Analysis
Clarity8/10
Novelty8/10
Utility9/10
Completeness8/10
Maintainability7/10
Pros & Cons

Pros

  • Streamlines the development process with AI assistance
  • Reduces the need for extensive documentation
  • Facilitates autonomous task execution
  • Enhances collaboration among team members

Cons

  • Requires familiarity with command-line tools
  • May have a learning curve for new users
  • Dependency on external tools like Claude Code
  • Potential for configuration errors if not validated

Related Skills

pytorch

S
toolCode Lib
92/ 100

“It's the Swiss Army knife of deep learning, but good luck figuring out which of the 47 installation methods is the one that won't break your system.”

View Analysis

agno

S
toolCode Lib
90/ 100

“It promises to be the Kubernetes for agents, but let's see if developers have the patience to learn yet another orchestration layer.”

View Analysis

nuxt-skills

S
toolCo-Pilot
90/ 100

“It's essentially a well-organized cheat sheet that turns your AI assistant into a Nuxt framework parrot.”

View Analysis

Disclaimer: This content is sourced from GitHub open source projects for display and rating purposes only.

Copyright belongs to the original author AnsarUllahAnasZ360.