A specialized development environment for transforming feature ideas into comprehensive specifications through a structured, spec-driven workflow. This project implements a three-phase methodology that takes rough concepts through requirements engineering, system design, and implementation planning.

This project provides a Claude Code agent specialized in Spec-Driven Agentic Development - a methodology that ensures complex features are properly planned before implementation begins. The agent guides you through a structured workflow that transforms ideas into actionable code.

The development process follows a strict three-phase iterative workflow:

• Creates user stories with acceptance criteria in EARS format
• Establishes clear functional and non-functional requirements
• Requires explicit user approval before proceeding

• Conducts necessary research and builds context
• Creates comprehensive design documentation with architecture, components, and data models
• Includes Mermaid diagrams for visual representations
• Requires explicit user approval before proceeding

• Converts design into actionable coding tasks
• Creates test-driven development approach with incremental progress
• Generates numbered checkbox task lists with requirement traceability
• Requires explicit user approval before execution begins

The project includes four specialized Claude Code slash commands:

Initializes a new feature specification by creating requirements, design, and task documents in the .claude/specs/{feature-name}/ directory structure.

Usage: /add-feature user authentication system

Analyzes your repository and creates comprehensive steering documents that give Claude persistent knowledge about your project:

• product.md - Product overview, features, and target audience
• tech.md - Technical stack, development guidelines, and best practices
• structure.md - Project structure, architecture patterns, and organization

Lists all feature directories in the .claude/specs/ directory, showing you what specifications have been created.

Executes specific tasks from a completed feature specification. The agent will:

• Read all specification files for context
• Execute ONE task at a time
• Mark tasks as in-progress (~) and completed (x)
• Wait for explicit approval before proceeding to next tasks

.claude/
├── CLAUDE.md                    # Main agent configuration and methodology
├── specs/                       # Feature specifications directory
│   └── {feature-name}/         # Individual feature directories
│       ├── requirements.md     # User stories and acceptance criteria
│       ├── design.md          # System design and architecture
│       └── tasks.md           # Implementation task list
├── steering/                   # Project context documents (created by command)
│   ├── product.md
│   ├── tech.md
│   └── structure.md
└── commands/                   # Slash command definitions
    ├── add-feature.md
    ├── create-steering-docs.md
    ├── list-feature-names.md
    └── start-task.md

• Sequential Workflow: Enforces Requirements → Design → Tasks progression
• User Approval Gates: Requires explicit approval between each phase
• EARS Format: Uses Event-driven Acceptance Requirements Specification format
• Test-Driven Planning: Creates implementation plans focused on testing and incremental development
• Requirement Traceability: Every task references specific requirements
• Single Task Execution: Prevents overwhelming complexity by focusing on one task at a time

You have two options for using these Claude Code slash commands:

Copy the .claude folder from this repository to your specific project:

# Clone or download this repository
git clone https://github.com/AshExplained/ccspecdev.git

# Copy the .claude folder to your project
cp -r ccspecdev/.claude /path/to/your/project/

This makes the slash commands available only within that specific project.

Copy the .claude folder to your home directory to make the slash commands available across all your projects:

# Clone or download this repository
git clone https://github.com/AshExplained/ccspecdev.git

# Copy the .claude folder to your home directory
cp -r ccspecdev/.claude ~/

This makes the slash commands accessible from any Claude Code session on your system.

1. Initialize a Feature: Use /add-feature your-feature-idea to start the specification process
2. Review & Approve: Go through each phase, providing feedback and approval
3. Execute Tasks: Use /start-task feature-name to begin implementation
4. Track Progress: Tasks are marked with status indicators (not started, in-progress ~, completed x)

• Never skip phases - must complete Requirements → Design → Tasks in order
• Always get explicit approval - clear "yes", "approved", or "looks good" required
• One task at a time - no automatic progression to next tasks
• Code focus only - tasks involve writing, modifying, or testing code exclusively
• Spec separation - this workflow creates planning artifacts, not production code

User: /add-feature user authentication system
→ Agent creates requirements.md with user stories
→ User approves requirements
→ Agent creates design.md with architecture
→ User approves design  
→ Agent creates tasks.md with implementation plan
→ User approves tasks

User: /start-task user-authentication
→ Agent executes first task only
→ User approves implementation
→ Agent marks task complete, waits for next instruction

This methodology ensures that complex features are properly planned, all stakeholders understand the requirements, and implementation follows a systematic, test-driven approach.