pimzino/claude-code-spec-workflow

This package adds structured development workflows to Claude Code, which is an AI coding assistant made by Anthropic. The idea is to give developers a predictable process for both building new features and fixing bugs, rather than jumping straight into writing code without a plan. For new features, the package provides a spec workflow that moves through four stages in order: writing requirements, creating a technical design, breaking the work into tasks, and then implementing those tasks. A single command starts the whole process, and the system generates additional commands automatically so each task can be executed on its own. For bug fixes, there is a separate workflow that moves through documenting the bug, analyzing the root cause, applying a fix, and then verifying that the fix worked. Once installed globally via npm and set up in a project directory, the package adds ten slash commands, four AI agents, and a set of document templates to the project's Claude Code configuration. It also creates three steering documents that capture the project's purpose, technology choices, and code organization rules. These documents give the AI persistent context about the project so it does not have to rediscover that information on every interaction. A real-time dashboard is included and can be launched from the command line to show the current status of specs and tasks. The dashboard can also be shared with external stakeholders through a temporary public URL using a tunnel feature, with optional password protection. It uses WebSocket connections to update live and integrates with git to track progress. The README notes that the primary development focus has moved to a separate MCP (Model Context Protocol) version of the same system, which is designed to work with a broader range of AI tools. This Claude Code-specific version remains available but is described as receiving limited updates going forward.