Domain Modeling

Degree of freedom: MIXED. Which term is canonical [HIGH freedom]; CONTEXT.md shape, lazy create, and the three ADR gates [LOW freedom — run exactly].

A shared language between user, agent, and codebase pays off every session:
consistent naming, fewer tokens spent describing concepts, and a codebase the
agent can navigate by vocabulary. Adapted from
mattpocock/skills (MIT).

This is the active discipline — challenging terms, inventing edge-case scenarios, and writing the glossary and decisions down the moment they crystallize. Merely reading CONTEXT.md for vocabulary is a one-line habit any skill can do; this skill is for changing the model, not consuming it.

How to reason

1. Observe — the word just used, and what CONTEXT.md already pins
2. Interpret — conflict, overload, or a genuine new concept
3. Classify — challenge / sharpen / write the glossary entry / offer an ADR
4. Capture — write the moment it crystallizes; do not batch

Worked example

Observe: user says "account" for both the billing Customer and the login User; CONTEXT.md is empty; accounts and auth.users both exist.
Interpret: one word, two entities — the next session will keep writing the wrong one.
Classify: sharpen now — Customer (billing party) vs User (login). Not an ADR: naming is still reversible.
Write immediately: Customer: the billing party. Avoid: account, user

Self-critique before reporting

• Domain language — definitions are not table or column names
• Avoid filled — rejected synonyms are listed
• Inline, not batched — each resolved term is in CONTEXT.md this turn
• Right owner — INDEX.md / rejected alternatives → docs-adr; interview-the-user → workflow-grilling

File structure [LOW freedom — this shape]

/
├── CONTEXT.md            ← the glossary (create on first resolved term)
├── docs/adr/             ← architectural decision records
│   └── 0001-<dash-case-title>.md
└── src/

Create files lazily — only when there is something to write. CONTEXT.md is a glossary and nothing else: no implementation details, no spec content, no scratch notes.

CONTEXT.md entry format

**<Canonical term>**:
<One- or two-line definition in domain language, not implementation language.>
_Avoid_: <rejected synonyms, comma-separated>

During the session [HIGH freedom]

• Challenge against the glossary. When the user uses a term that conflicts

with 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 a term is vague or overloaded, propose a

precise canonical term: "You're saying 'account' — do you mean the Customer or the User?"

• Stress-test with concrete scenarios. Invent edge-case scenarios that force

precision about the boundaries between concepts.

• Cross-reference with code. When the user states how something works, check

whether the code agrees, and surface contradictions.

• Update CONTEXT.md inline. Capture each resolved term the moment it is

resolved — don't batch.

ADRs — offer sparingly [LOW freedom — all three gates]

Only offer to create an ADR when all three hold:

1. Hard to reverse — changing your mind later costs something real.
2. Surprising without context — a future reader will ask "why this way?"
3. A real trade-off — genuine alternatives existed and one was picked for

specific reasons.

ADR body: Status, Context (the forces), Decision (what and why), Consequences (what becomes easier/harder). Number sequentially in docs/adr/.

Related: workflow-grilling (use both in one session to interview and capture the language), design-prd, docs-coauthor.