| layout | title | nav_order | permalink |
|---|---|---|---|
default |
Home |
1 |
/ |
A framework for building LLM coding guidelines that actually work.
Contributing to the documentation? Start with the Contributing Guide. It defines the issue process, PR workflow, and quality bar we use to keep this framework strong.
Most teams using LLMs for code generation struggle with the same problem: the model doesn't know your project. It doesn't know your stack, your conventions, your architecture decisions, or what it should never touch. So it guesses — and the guesses are inconsistent, sometimes wrong, and expensive to fix.
The fix isn't better prompts. It's better project context.
guide-LLiMes teaches you why each type of guideline matters and how to write one that LLMs can actually follow. It's not a template to copy-paste. It's a method for building guidelines that fit your specific project, team, and tools.
- Team leads and architects setting up guidelines for a team
- Individual developers who want better LLM output on their own projects
- Anyone using Claude Code, Cursor, GitHub Copilot, or other LLM coding tools
Not all guidelines are equally important. guide-LLiMes organizes them into three tiers:
| Tier | Purpose | When to add |
|---|---|---|
| Must-have | Prevent breakage. Without these, LLMs will hallucinate features, break APIs, and create files in wrong places. | Stage 1 |
| Should-have | Reduce variance. These make LLM output more consistent and reviewable. | Stage 2 |
| Nice-to-have | Scale and compliance. These matter when your team grows or regulations apply. | Stage 3+ |
Start with 3 must-haves. Add more as you see what the LLM gets wrong.
| You want to... | Start here | Time |
|---|---|---|
| Get guidelines running fast | Three-tier model → pick 3 from Must-haves → write them | ~15 min |
| Understand the full framework | Why guidelines matter → read through Fundamentals → all Must-haves | ~2 hrs |
| Set up a specific tool | Tool guides → your tool → reference Must-haves for content | ~30 min |
| Understand LLM context limits | Token budgets | ~10 min |
If you want to improve this framework, use the Contributing Guide before opening a PR. We intentionally ask contributors to invest in the core concepts first so new content raises the quality bar instead of adding noise.
- Introduction
- Fundamentals
- Must-Have Guidelines
- Should-Have Guidelines
- Nice-to-Have Guidelines
- Tool Guides
- Glossary
- Contributing Guide
