shinpr/claude-code-workflows

Production-ready development workflows for Claude Code, powered by specialized AI agents.

What it solves

Claude Code Workflows provides a structured, agentic development lifecycle for Claude Code, ensuring that generated code is strictly aligned with design documents and tests. It prevents the common issue of AI-generated code drifting from the original requirements or failing to pass its own tests by introducing specialized agents for each stage of the development process.

How it works

The project implements a series of "recipes" (entry points) that route user requests through a pipeline of specialized agents. Depending on the task complexity (small, medium, or large), the workflow varies:

  • Analysis & Planning: A requirement-analyzer determines the scale of the work. For larger tasks, it triggers the creation of Product Requirement Documents (PRDs), codebase analysis, and technical design docs.
  • Execution: Specialized executors (backend or frontend) implement the feature using Test-Driven Development (TDD), ensuring code is type-checked and linted.
  • Verification: A code-reviewer and code-verifier check the implementation against the design docs to ensure no requirements were missed.
  • Diagnosis: A dedicated loop between an investigator, verifier, and solver is used to root-cause bugs and propose trade-off analyzed solutions.
  • Reverse Engineering: The system can scan existing code to generate PRDs and Design Docs, making legacy codebases AI-friendly.

Who it’s for

Developers using Claude Code who want a more deterministic, professional software engineering process rather than simple chat-based code generation. It is particularly useful for those working on backend APIs, React/TypeScript frontends, or full-stack applications.

Highlights

  • Layer-Aware Workflows: Separate plugins for backend, frontend, and full-stack development with specialized agents for each (e.g., ui-spec-designer for React).
  • Multi-Agent Orchestration: Uses a chain of agents (Analyzer $\rightarrow$ Designer $\rightarrow$ Planner $\rightarrow$ Executor $\rightarrow$ Reviewer) to maintain a clean context window.
  • UI Specification: A dedicated step for frontend work that captures edge cases (loading, error, empty states) often missed by prototypes.
  • Automated Documentation: Ability to reverse-engineer existing code into structured PRDs and design docs.
  • TDD Integration: Built-in principles for test-first implementation and quality fixing.