4meta5

project-init

0
0
# Install this skill:
npx skills add 4meta5/skills --skill "project-init"

Install specific skill from multi-skill repository

# Description

|

# SKILL.md


name: project-init
description: |
Scaffold new projects with standard structure. Use when: (1) user says
"create project", "scaffold", "initialize", (2) starting from scratch,
(3) need consistent project layout. Creates CLAUDE.md, README.md, PLAN.md,
RESEARCH.md, AGENTS.md, .gitignore, and .claude/skills/ directory.
category: development
user-invocable: true


Project Initialization

Scaffolds new projects with a standard structure for Claude Code workflows.

Trigger Conditions

Invoke when user says:
- "create a new project"
- "scaffold a project"
- "initialize project"
- "start new project"
- "set up project structure"

Procedure

Step 1: Determine Location

Check if currently inside a git workspace:

git rev-parse --is-inside-work-tree 2>/dev/null

If inside git workspace:
- Navigate to parent directory
- Create new directory there
- This prevents nested git repos

If not inside git workspace:
- Create in current directory

Step 2: Create Directory Structure

mkdir -p {{project_name}}
cd {{project_name}}
mkdir -p .claude/skills
mkdir -p src tests docs

Step 3: Initialize Git

git init

Step 4: Create Standard Files

Create these files from templates:

File Purpose
CLAUDE.md Project guidance for Claude
README.md User-facing documentation
PLAN.md Remaining work tracker
RESEARCH.md Investigation notes
AGENTS.md Agent coordination
.gitignore Comprehensive patterns

Step 5: Create .gitignore

# Dependencies
node_modules/
.pnpm-store/

# Build outputs
dist/
build/
.svelte-kit/
.next/
.nuxt/

# Test artifacts
coverage/
test-results/
playwright-report/
.nyc_output/

# IDE
.idea/
.vscode/
*.swp
*.swo

# OS
.DS_Store
Thumbs.db

# Environment
.env
.env.local
.env.*.local

# Logs
*.log
npm-debug.log*
yarn-debug.log*
yarn-error.log*

# Cache
.cache/
.turbo/
.eslintcache
.parcel-cache/

# Credentials
*.pem
*.key
credentials.json

Step 6: Populate Templates

Fill templates with project-specific values:

CLAUDE.md:
- Project name
- Detected language (TypeScript, Python, etc.)
- Project structure
- Recommended skills based on tech stack

README.md:
- Project name and description
- Quick start commands
- Basic usage example

PLAN.md:
- Initial tasks based on user's goals
- Empty backlog section
- Empty completed section

RESEARCH.md:
- Topic placeholder
- Empty findings section

AGENTS.md:
- Empty, ready for agent coordination

Step 7: Initial Commit

git add .
git commit -m "Initial project scaffold

Created standard structure with:
- CLAUDE.md (project guidance)
- README.md (documentation)
- PLAN.md (work tracker)
- RESEARCH.md (investigation notes)
- AGENTS.md (agent coordination)
- .gitignore (comprehensive patterns)

Co-Authored-By: Claude <[email protected]>"

Output Format

After scaffolding, report:

Project scaffolded: {{project_name}}

Created files:
  - CLAUDE.md
  - README.md
  - PLAN.md
  - RESEARCH.md
  - AGENTS.md
  - .gitignore
  - .claude/skills/ (directory)

Next steps:
1. Review CLAUDE.md and customize for your project
2. Update README.md with project description
3. Add initial tasks to PLAN.md
4. Install skills with: skills scan --all

Customization Options

User can specify:
- --language: Primary language (typescript, python, rust, go)
- --framework: Framework (react, svelte, next, sveltekit)
- --no-git: Skip git initialization
- --skills: Comma-separated list of skills to install

Example Usage

Basic:

/project-init my-app

With options:

/project-init my-app --language typescript --framework sveltekit

Skill Chaining

After scaffolding:
1. dogfood-skills runs skills scan to recommend skills
2. doc-maintenance is ready to update PLAN.md
3. tdd is available for test-driven development

Notes

  • Does not overwrite existing files
  • Creates parent directories as needed
  • Uses Handlebars-style templates ({{variable}})
  • Respects user's existing git configuration

# Supported AI Coding Agents

This skill is compatible with the SKILL.md standard and works with all major AI coding agents:

Learn more about the SKILL.md standard and how to use these skills with your preferred AI coding agent.