Create Example

Read DEVELOPERS.md to see the available cast of characters. These developers will appear in the scenario narrative.

Present a concept summary to the user for approval before proceeding to Phase 2.

Phase 2: Schema Design → SCHEMA_DESIGN.md

Use EnterPlanMode to design the full layer structure before writing anything.

Layer design principles:

• Layer 0: Core tables with no dependencies (the "root" entities)
• Layer 1+: Build on previous layers; each layer adds a coherent capability
• Parallel layers (1a, 1b, 1c): Independent features that can be built in any order
• Dependencies converge: a later layer that requires all of 1a/1b/1c lists all three as dependencies
• Add intentional technical debt: one or two design mistakes that get corrected later (adds realism)
• Aim for 1–3 tables per layer (avoid monolithic layers)

Platform-agnostic spec conventions:

• Column types use logical descriptions, not platform SQL:
  • Use identifier, auto-generated not INT AUTO_INCREMENT
  • Use text, up to 100 characters not VARCHAR(100)
  • Use boolean not TINYINT(1) or BIT
  • Use datetime not DATETIME or DATETIME2
  • Use decimal(10,2) format for decimals
• Mark required/optional explicitly: required / optional, nullable
• Note default values where they exist
• Note indexes on non-PK columns that will be needed

Status markers:

• 📋 Planned — not yet implemented in any implementation
• ✅ Complete — implemented in at least one implementation
• ⚠️ Blocked — depends on something not yet done

Read docs/SCHEMA_AND_SCENARIO_GUIDE.md for the full authoring spec before writing.

File location: <Scenario-DB>/SCHEMA_DESIGN.md (at the scenario root, not inside any implementation folder)

After writing SCHEMA_DESIGN.md, present it to the user for review before proceeding to Phase 3.

Phase 3: Scenario Writing → SCENARIO.md

Write the team narrative that gives the schema history and context.

Narrative elements:

• Pick 3–5 developers from DEVELOPERS.md who will "build" this database
• Assign roles: who leads, who implements, who reviews
• Set a realistic start date (past, not future) and a timeline spanning weeks or months
• Each layer has a date/time it was implemented — these become patch folder timestamps
• Include realistic details: design discussions, mistakes made, corrections, reviews
• Write developer conversations (brief exchanges, not essays)
• Make technical debt explicit: show the decision that caused it, and the later correction

Timestamp rules:

• Work happens during business hours (9am–6pm)
• Developers work in bursts with gaps (not one layer per hour all day)
• Multiple layers on one day is fine; a few weeks between phases is realistic
• Timestamps must be consistent with the dependency order in SCHEMA_DESIGN.md

Read docs/SCHEMA_AND_SCENARIO_GUIDE.md for the full scenario authoring spec and DEVELOPERS.md for developer profiles.

File location: <Scenario-DB>/SCENARIO.md (same folder as SCHEMA_DESIGN.md)

Phase 4: Scaffold the Folder Structure

After the user approves SCENARIO.md, create the initial folder structure:

mkdir -p <Scenario-DB>/test-data

Stub test-data/data-manifest.json:

{
  "loadOrder": [],
  "tables": {}
}

Leave loadOrder and tables empty — they will be populated when test data CSVs are created (Step 6b of the project plan).

Do NOT create any implementation folder yet — that is the new-implementation skill's job.

Phase 5: Update README.md

Add the new scenario to the scenario index table in the repo root README.md:

• Add a row to the "Scenarios" table with: scenario name, description, layer count (from SCHEMA_DESIGN.md), status (In progress), implementations (None yet)
• Do not rewrite other sections of README.md

Phase 6: Summary

Report to the user:

• Scenario name and folder created
• Layer count and structure summary
• Developers assigned from DEVELOPERS.md
• Timeline span
• Files created: SCHEMA_DESIGN.md, SCENARIO.md, test-data/data-manifest.json
• Next step: use /new-implementation to scaffold the first implementation

Important Reminders

• Both canonical docs are shared across all implementations — never add platform-specific content to them
• Get user approval at the end of Phase 1 (concept) and Phase 2 (schema) before continuing
• Timestamps in SCENARIO.md drive patch folder names — make them realistic and consistent with the dependency order
• Read DEVELOPERS.md before assigning characters — use the defined profiles, don't invent new developers