Claude Workspace Initialization Skill

Starting 2026 with a new technical approach. This resource describes a comprehensive methodology for initializing Claude Code workspaces efficiently and scalably.

🤔 The Problem

When working with Coding Agents in different projects/repositories that don't yet have Agents, Skills, Commands, Hooks, etc., I need to run /init (with Claude Code) or similar commands in each repository to generate the CLAUDE.md or AGENTS.md.

After that base file, comes the complex part: building the architecture of Agents, Skills, Commands, Hooks, etc., for each specific repository.

Additional Challenges

• Multiple repositories: A product usually involves more than one repository (front, back, db, infra, etc.)
• Specific technologies: Each component has its own technologies and frameworks
• Custom configuration: Each file should be designed specifically for its project
• Consistency: Maintaining coherent patterns between related repositories

💡 The Solution: Global Skill + Workspace

The solution combines two Claude Code capabilities:

1. Claude Code Workspace

In each Claude Code session you can run /add-dir and pass the absolute (or relative) path of another repository. This allows a single session to modify multiple repositories simultaneously.

It's the same Workspace concept as any IDE, but applied to Claude Code per session.

2. Global Skill

Claude Code has a folder hierarchy for defining Agents, Skills, Hooks, etc. By creating a Skill in ~/.claude/skills, that Skill becomes globally available.

🛠️ The Methodology

Automatic Analysis

The skill automatically analyzes each project to identify:

• Exact technologies: Specific versions of languages, frameworks and runtime
• Dependencies: Complete catalog of libraries from package.json, pom.xml, requirements.txt
• Architecture: Structure, layers, patterns and existing conventions
• Quality tools: Linters, formatters, static analyzers
• Git conventions: Existing commit style

Complete Configuration

For each project, the skill creates:

Base Documentation

• CLAUDE.md: Claude Code specific context
• AGENTS.md: Agnostic context following the open standard

Specialized Skills

• Software Engineering: SOLID principles, Clean Code with project examples
• Language/Framework Expert: Specific features of the exact version
• Library-specific: Skills for main libraries (UI, charting, AI/ML, etc.)
• SDLC: Git workflow, quality tools, review process

Intelligent Agents

• Technology Experts: One per main language with full tool access
• Framework Experts: Specialized in specific frameworks
• Design Patterns: Patterns actually used in the project
• Code Reviewer: Read-only, organized by severity

Useful Commands

• /commit: Linting + review + conventional commit
• /lint: Project-specific tools
• /review: Review by severity
• /build and /run: Stack-specific commands

🎯 Key Features

Technology Specificity

• Each agent and skill is specific to the project's technology
• Code examples that match real conventions
• Explicit version constraints (what's allowed vs what's not)

Scalability

• Works with individual project or multiple repositories
• Parallel analysis using subagents
• Preserved configuration for existing settings

Guaranteed Quality

• Applicable design patterns, not theoretical
• Exact framework versions (Spring Boot 2.x vs 3.x, Angular 8 vs 17)
• Real project conventions (indentation, quotes, naming)

📄 The Complete Skill

The downloadable file contains the complete skill prompt I've been using these past months. It includes:

• Detailed instructions for each phase of the process
• Templates for all file types
• Specific examples for different technology stacks
• Constraints and best practices to guarantee quality
• Summary matrix for configuration tracking

🚀 Practical Usage

Installation

1. Download the SKILL.md file
2. Place it in ~/.claude/skills/init-workspace/SKILL.md
3. The skill will be globally available in all sessions

Execution

# For all projects in the workspace
/init-workspace

# For a specific project
/init-workspace /path/to/specific/project

Result

• Complete Claude Code configuration for each project
• Specialized agents ready to use
• Custom commands for common workflows
• Updated technical documentation

🔄 Replicability

This methodology can be replicated in any CLI or AI IDE:

• Cursor: Very simple using commands
• Codex: Adaptable to its skill system
• Gemini CLI: Compatible with its configuration structure

💼 Use Cases

New Projects

• Complete initial configuration from scratch
• Establishment of patterns and conventions
• Quality tools setup

Existing Projects

• Migration to Claude Code
• Update of obsolete configurations
• Standardization between repositories

Development Teams

• Quick onboarding of new members
• Consistency in configurations
• Automated best practices

This skill represents months of refinement and real usage across multiple projects, incorporating lessons learned and optimizations based on practical experience.