Domain Glossary

Why This Matters

Without a shared vocabulary, the LLM invents its own. One generation calls it a “customer,” the next calls it a “user,” the third calls it a “client.” The model doesn’t know these mean different things in your domain — or that they mean the same thing.

A domain glossary establishes the ubiquitous language for your project. Every entity, concept, and action has one name.

What to Include

  • Core entities and their definitions — the nouns of your domain
  • Relationships — how entities relate to each other
  • Terms to avoid — synonyms that should not be used
  • Business invariants — rules that must always hold true

Example

## Domain Glossary

Workspace — the top-level organizational unit. One workspace has many projects.
  NOT: "organization", "team", "account"

Project — a collection of tasks with a lifecycle (active, archived).
  NOT: "board", "sprint", "group"

Member — a user associated with a workspace. Has a role (owner, editor, viewer).
  NOT: "participant", "collaborator", "teammate"

Task — a unit of work within a project. Has status (todo, in_progress, done).
  NOT: "ticket", "issue", "item", "card"

Invariants:
- A workspace must always have at least one owner
- A task always belongs to exactly one project
- Archived projects cannot have active tasks

Common Mistakes

Only defining terms, not banning synonyms. The model’s biggest naming mistake is using a plausible synonym. Explicitly listing what NOT to call things is essential.

Too long. A glossary with 50 entries consumes too many tokens. Focus on the 10-15 terms that the model confuses most often.

Not including invariants. Domain rules like “a workspace must have an owner” prevent the model from generating code that violates business logic.

Tool-Specific Notes

  • Claude Code: Include key glossary terms in CLAUDE.md. Full glossary can be in a separate file the model reads on demand.
  • Cursor: Good candidate for an alwaysApply: true rule if domain consistency is a frequent problem.
  • All tools: Most impactful for domain-rich applications (healthcare, finance, e-commerce) where terminology is precise and consequential.

References


Back to top

guide-LLiMes — Build LLM coding guidelines that work.

This site uses Just the Docs, a documentation theme for Jekyll.