Useful Coding Guides For LLMs — AI skill for Claude Code

Useful Coding Guides for LLMs

This repository contains a growing body of documentation about how to build robust, maintainable Next.js 15 applications while collaborating with large language model coding tools. The goal is to provide a clear reference so that both humans and AI assistants can work together on complex codebases.

The main documents in this repository include:

• NEXTJS15_BEST_PRACTICES.md – a nearly 1000 line guide describing recommended architecture and patterns for modern Next.js 15 projects.
• NEXT_BEST_PRACTICES_IMPLEMENTATION_PROGRESS.md – a detailed checklist that tracks which parts of the guide have been implemented in a real codebase.
• PROMPT_TO_GENERATE_AND_REVISE_GUIDES.md – the seed prompt used to research best practices and compile the guide.
• EXAMPLE_USING_THE_TECHNIQUE_IN_CLAUDE_CODE.md – a transcript demonstrating how to iteratively apply the guide with an agentic coding assistant like Claude Code.

Recommended Workflow

The documentation in this repo grew out of a real project that combined a Python FastAPI backend with a modern Next.js 15 frontend. To establish a solid baseline, multiple models — Claude Opus 4, OpenAI’s GPT‑4o, and Gemini Deep Research — were asked to scour the web for the latest best practices in this stack. Their findings were merged into a single guide using a model with a large context window (Gemini 2.5 Pro or Opus works well).

Iterate on the guide until the models largely agree that it captures the current state of the art. Example prompts for this refinement loop live in `PROMPT_TO_GENERATE_AND_REVISE_GUIDES.md`, and the resulting document is `NEXTJS15_BEST_PRACTICES.md`.

Once the guide is stable, commit your code to git and launch Claude Code in background task mode. Many users create a convenient alias:

alias cc='ENABLE_BACKGROUND_TASKS=1 claude --dangerously-skip-permissions'

Ask Claude Code to carefully read the best practices guide and begin implementing it acro