domain-model

Legacy: prefer /grill-with-docs. Keep only for backward compatibility with local three-hat workflow.

Domain Model

Grill + document. Three-hat adversarial review PLUS institutional memory captured inline as CONTEXT.md + ADRs.

Step 1: Interview

Walk every branch of decision tree. Resolve dependencies one-by-one. Each question, give recommended answer. One at time.

If question answerable by exploring codebase, explore first.

Light DDD -- Document, Don't Prescribe

USE: Ubiquitous Language | Bounded Contexts | ADRs SKIP: Entities | Value Objects | Aggregates | Domain Events

Goal = "just enough docs" make codebase navigable. Language into software, not patterns into software.

Domain Awareness

During codebase exploration, look for existing docs:

Single context (most repos)

/
├── CONTEXT.md
├── docs/adr/
│   ├── 0001-event-sourced-orders.md
│   └── 0002-postgres-for-write-model.md
└── src/

Multi-context (if CONTEXT-MAP.md exists at root)

/
├── CONTEXT-MAP.md
├── docs/adr/                    <- system-wide decisions
├── src/
│   ├── ordering/
│   │   ├── CONTEXT.md
│   │   └── docs/adr/            <- context-specific decisions
│   └── billing/
│       ├── CONTEXT.md
│       └── docs/adr/

Create files lazy -- only when first term resolved or first ADR needed.

Step 2: Domain Grill

Challenge glossary: Term conflict with CONTEXT.md? Call out now.

Sharpen language: Vague or overloaded term? Propose precise canonical term. "You say 'account' -- Customer or User? Different things."

Concrete scenarios: Stress-test relationships with edge cases. Force precision on boundaries.

Cross-reference code: User state how something works -> verify code agrees. Surface contradictions.

Step 3: Three-Hat Fan-Out (parallel)

Once user present coherent plan + terms resolved, spawn three reviewers in parallel (single message, multiple Agent tool calls):

• plan-product-hat: persona, pain, success metric, scope, reversibility, TTV
• plan-engineering-hat: architecture, error paths, perf, security, test strategy, rollback
• plan-design-hat: flow, a11y, copy, visual consistency, states (empty/loading/error)

Each emit {reviewer, status, findings[], must_answer[]} per findings-schema.md.

Step 4: Merge + Document

Consolidate all must_answer questions into single list, deduped. Surface to user. User answer each. Plan updated inline.

Update CONTEXT.md inline: Term resolved or scope clarified -> update now. No batch. See CONTEXT-FORMAT.md.

Offer ADRs sparingly: Only when ALL three true: hard reverse, surprising without context, result of real trade-off. See ADR-FORMAT.md.

Any reviewer return status: BLOCKED -> plan no advance until blocking finding addressed or user override.

Step 5: Approve

All hats return APPROVED or user override specific findings -> plan approved -> Phase 3 (Implement).

Skip Gate

Skip three-hat fan-out only if:

• Trivial bug fix, AND
• <3 tasks in plan, AND
• No architectural / product / UX decisions

Else, fan-out mandatory. For a light grill without DDD docs, use /grill-me instead.