Product Introduction
Definition: /dev for Claude Code is an open-source, multi-agent development Standard Operating Procedure (SOP) designed as a custom skill for the Anthropic Claude Code CLI. It functions as an orchestration layer that transforms a standard AI coding assistant into a structured "Tech Lead" capable of managing parallel "Worker Agents" across isolated git worktrees.
Core Value Proposition: While traditional AI coding tools focus on autocomplete and chat-based code generation, /dev for Claude Code provides a comprehensive engineering framework. It exists to solve the "chaos of scale" in AI-generated code by enforcing a 6-phase software development lifecycle (SDLC). By shifting Claude from a "coder" to a "Tech Lead," the tool ensures architectural integrity, prevents merge conflicts through pre-coding scans, and mandates security audits, making AI-driven development viable for complex, multi-feature projects.
Main Features
Six-Phase Structured Engineering SOP: The core of the product is a rigid workflow that guides every task through a defined lifecycle. This begins with Phase 0 (Request Classification), followed by Phase 1 (Product Requirement Document Alignment), Phase 2 (Architecture and Task Decomposition), Phase 3 (Parallel Execution), Phase 4 (Structured Code Review), and Phase 5 (Retrospective). This systematic approach prevents the "context drift" common in long AI chat sessions by requiring explicit sign-offs at each milestone.
Parallel Multi-Agent Orchestration via Git Worktrees: /dev leverages git worktrees to enable the "Worker Agent" paradigm. When a complex task is decomposed, the Tech Lead spawns multiple sub-agents, each operating in an isolated file-system directory (worktree). This allows for concurrent development of frontend components, backend APIs, and database schemas without file-system interference, drastically reducing the time-to-merge for multi-module features.
Automated Security and Quality Gates: Unlike standard AI prompts, /dev integrates native CLI tools to enforce code quality. The system runs mandatory security audits using bandit and pip-audit for Python or npm audit for JavaScript environments. Additionally, it implements a 6-category "counterexample check" (Null, Empty, Boundary, External Failure, Concurrency, and Malicious Input) and a Database Migration Guard that triggers a veto on dangerous DDL operations like ALTER TABLE or DROP COLUMN without manual approval.
Problems Solved
Pain Point: AI-Generated Code Regressions and Conflicts. Standard LLM interactions often lead to overlapping changes that break existing functionality. /dev solves this with a pre-coding conflict scan that checks open GitHub Issues and Pull Requests for file overlaps before a Worker Agent begins writing code.
Target Audience: This tool is built for Professional Software Engineers, Tech Leads, and Full-Stack Developers who use the Claude Code CLI. It is particularly valuable for Senior Developers who want to delegate implementation details to AI while maintaining control over architecture and security standards.
Use Cases:
- Bootstrapping New Projects: Rapidly turning a vague idea into a PRD, architecture map, and initial codebase.
- Parallel Feature Development: Building multiple related modules (e.g., Auth, CRUD, and UI) simultaneously in separate worktrees.
- Systematic Refactoring: Executing large-scale code changes that require strict adherence to a pre-defined architectural plan and mandatory security verification.
Unique Advantages
Differentiation from Autocomplete Tools: Most AI tools (like GitHub Copilot or Cursor) operate at the line-of-code or file level. /dev for Claude Code operates at the process level. It defines the "Who, What, and How" of the development cycle, ensuring that the AI never writes code directly in the main conversation, which preserves the "Tech Lead" persona and prevents architectural degradation.
Key Innovation: The "Main Conversation Isolation" rule. By strictly forbidding the main Claude instance from writing code and forcing all changes through specialized Worker Agents and Pull Requests, /dev maintains a clean project context. This approach mimics a high-performing human engineering team, where the Lead provides direction and the Workers provide implementation, separated by a rigorous review process.
Frequently Asked Questions (FAQ)
How does /dev for Claude Code manage merge conflicts in parallel tasks? /dev uses a proactive conflict detection mechanism. Before a Worker Agent is permitted to start a task in a new worktree, the system scans all currently open GitHub Issues and active branches for overlapping file modifications. If a potential conflict is detected, the Tech Lead alerts the user to sequence the tasks or merge existing changes before proceeding.
What are the system requirements for installing the /dev skill? To use /dev, users must have the official Claude Code CLI (Anthropic) installed and authenticated. Additionally, the GitHub CLI (gh) is required for managing Issues and Pull Requests, and a modern version of Git must be installed to support the worktree functionality used for parallel agent execution.
Does /dev for Claude Code support both Windows and macOS? Yes. The project provides native installation scripts for both environments: a bash script (install.sh) for macOS and Linux users, and a PowerShell script (install.ps1) for Windows users. It also supports both English and Chinese languages for the development workflow and documentation.
