Domain Modeling

Domain Modeling

Actively build and sharpen the project's domain model as you design — the active discipline of challenging terms, inventing edge-case scenarios, and writing the glossary down the moment it crystallises. Merely reading CONTEXT.md for vocabulary is not this skill (that's a one-line habit any skill can do); this skill is for when you're changing the model, not just consuming it.

Why it pays off: an opinionated CONTEXT.md is a persistent, shared language. Once the domain has canonical names, variables/functions/files get named consistently, the codebase is easier for an agent to navigate, and the agent spends fewer tokens on thinking because it has a more concise language. The cost is one glossary entry; the payoff compounds every session.

Adapted from mattpocock/skills (MIT) — the domain-modeling skill, with ADR-recording delegated to this repo's adr skill.

File structure

Most repos have a single context — one CONTEXT.md at the repo root:

/
├── CONTEXT.md          ← the glossary
└── src/

If a CONTEXT-MAP.md exists at the root, the repo has multiple contexts; the map points to where each CONTEXT.md lives (e.g. src/ordering/CONTEXT.md, src/billing/CONTEXT.md) and how they relate. See CONTEXT-FORMAT.md for both shapes.

Create files lazily — only when you have something to write. If no CONTEXT.md exists, create one when the first term is resolved.

Wire it in for future sessions. The payoff — consistent naming and fewer tokens every session — only lands if CONTEXT.md is loaded when future sessions start, not just when this skill fires. When you create CONTEXT.md (or update one and notice it is unreferenced), check the repo's CLAUDE.md: if it does not reference CONTEXT.md, ask the user once whether to wire it in — @CONTEXT.md for a small glossary, or a "consult CONTEXT.md for domain terms" line for a larger one — and add it on yes (don't re-ask on no). If there is no CLAUDE.md at all, mention the option rather than creating one uninvited.

Decisions are not vocabulary. When a hard-to-reverse, non-obvious trade-off surfaces (not a term), record it as an ADR — use this repo's adr skill, which owns the ADR format (docs/adr/YYYY-MM-DD-<slug>.md). Do not invent a second ADR convention here.

During the session

Challenge against the glossary

When the user uses a term that conflicts with the existing language in CONTEXT.md, call it out immediately. "Your glossary defines 'cancellation' as X, but you seem to mean Y — which is it?"

Sharpen fuzzy language

When the user uses vague or overloaded terms, propose a precise canonical term. "You're saying 'account' — do you mean the Customer or the User? Those are different things."

Flag collisions with technical terms

When a chosen domain term shadows a well-known general-programming term likely to appear in the same codebase — "rollback" (vs. a database transaction rollback), "event", "handler", "session" — surface the collision even if the domain term is already precise. Offer a less-ambiguous canonical name; if the user keeps theirs, note the collision in the glossary entry so a future reader isn't misled. (This is distinct from challenging against the glossary, which is about conflicts with existing entries, and from sharpening fuzzy language, which is about vague business terms.)

Discuss concrete scenarios

When domain relationships are being discussed, stress-test them with specific invented scenarios that probe edge cases and force precision about the boundaries between concepts.

Cross-reference with code

When the user states how something works, check whether the code agrees. If you find a contradiction, surface it: "Your code cancels entire Orders, but you just said partial cancellation is possible — which is right?"

Update CONTEXT.md inline

When a term is resolved, update CONTEXT.md right there — don't batch these up, capture them as they happen. Use the format in CONTEXT-FORMAT.md.

CONTEXT.md is a glossary and nothing else — totally devoid of implementation details. It is not a spec, not a scratch pad, and not a home for implementation decisions (those are ADRs). Only include terms specific to this project's domain; general programming concepts (timeouts, error types, utility patterns) don't belong even if the project uses them heavily.

Relationship to other skills

• Decisions, not terms → adr. A load-bearing, hard-to-reverse trade-off is an ADR, not a glossary entry.
• Exploring intent → brainstorming. A grilling/brainstorming session is where terms surface; this skill is what captures them.
• Not architecture vocabulary. Module, interface, and seam are architecture language; CONTEXT.md is domain language. Keep them distinct.