backend-data-model

Generate production-quality data model documentation from entity descriptions.

Context Files

• $JAAN_CONTEXT_DIR/tech.md - Tech stack context (CRITICAL: determines database engine and patterns)
  • Uses sections: #current-stack, #constraints, #patterns
• $JAAN_CONTEXT_DIR/config.md - Project configuration
• $JAAN_TEMPLATES_DIR/jaan-to:backend-data-model.template.md - Output template
• $JAAN_LEARN_DIR/jaan-to:backend-data-model.learn.md - Past lessons (loaded in Pre-Execution)
• ${CLAUDE_PLUGIN_ROOT}/docs/extending/language-protocol.md - Language resolution protocol

Input

Entities: $ARGUMENTS

Accepts any of:

• Entity list — Comma-separated entity names (e.g., "User, Post, Comment")
• PRD reference — Path to PRD file with data requirements
• Existing schema — Path to DDL/migration file for enhancement
• Feature description — Free text describing the feature's data needs

If no input provided, ask: "What entities or features should the data model cover?"

---

Pre-Execution Protocol

MANDATORY — Read and execute ALL steps in: ${CLAUDE_PLUGIN_ROOT}/docs/extending/pre-execution-protocol.md
Skill name: backend-data-model
Execute: Step 0 (Init Guard) → A (Load Lessons) → B (Resolve Template) → C (Offer Template Seeding)

Also read tech context (CRITICAL for this skill):

• $JAAN_CONTEXT_DIR/tech.md - Determines database engine, constraints, common patterns

Language Settings

Read and apply language protocol: ${CLAUDE_PLUGIN_ROOT}/docs/extending/language-protocol.md
Override field for this skill: language_backend-data-model

Language exception: Generated code output (variable names, code blocks, schemas, SQL, API specs) is NOT affected by this setting and remains in the project's programming language.

---

PHASE 1: Analysis (Read-Only)

Thinking Mode

ultrathink

Use extended reasoning for:

• Extracting constraints from natural language (uniqueness, cardinality, CHECK)
• Mapping entity relationships and detecting implicit constraints
• Planning index strategy using ESR ordering
• Assessing migration complexity per table

Step 1: Parse Input

Analyze the provided input to extract entities:

If entity list:

1. Split comma-separated names
2. Infer relationships from naming (e.g., "Comment" implies parent "Post")
3. Note implied attributes (e.g., "User" implies email, name)

If PRD reference:

1. Read the PRD file
2. Extract data-relevant user stories and acceptance criteria
3. Identify entities (nouns), relationships (verbs), and constraints
4. Note technical constraints mentioned

If existing schema:

1. Read DDL/migration file
2. Extract tables, columns, types, constraints, indexes
3. Map foreign keys to relationships
4. Identify gaps or enhancement opportunities

If feature description:

1. Parse for entity nouns and relationship verbs
2. Identify implied constraints and data rules
3. Note any retention or compliance mentions

Build initial understanding:

INPUT SUMMARY
─────────────
Type:          {entity-list/prd/schema/description}
Entities:      {list of identified entities}
Relationships: {implied relationships}
Constraints:   {mentioned constraints}
Unknown:       {areas needing clarification}

Step 2: Clarify Data Design

Ask up to 6 smart questions based on what's unclear from Step 1. Skip questions already answered by the input or tech.md.

Engine question (ask if not in tech.md):

1. Use AskUserQuestion:
   • Question: "Which database engine(s) should the model target?"
   • Header: "Engine"
   • Options:
     • "PostgreSQL (Recommended)" — Full constraint support, JSONB, RLS, transactional DDL
     • "MySQL" — INSTANT DDL (8.0+), InnoDB clustered index, CHECK (8.0.16+)
     • "SQLite" — Lightweight, limited ALTER TABLE, edge/embedded use cases
     • "Multiple engines" — Generate engine-specific variants

Migration question (always ask):
2. Use AskUserQuestion:

• Question: "Is this a greenfield schema or extending existing tables?"
• Header: "Migration"
• Options:
  • "Greenfield (Recommended)" — New schema, CREATE TABLE statements
  • "Brownfield" — Extending existing tables, zero-downtime migration notes
  • "Mixed" — Some new tables, some alterations

Tenancy question (ask if not in tech.md constraints):
3. Use AskUserQuestion:

• Question: "Does this system require multi-tenancy?"
• Header: "Tenancy"
• Options:
  • "No multi-tenancy" — Single-tenant application
  • "Shared tables + tenant_id (Recommended)" — Discriminator column with RLS
  • "Schema-per-tenant" — Separate schemas per tenant
  • "Database-per-tenant" — Full isolation, regulated industries

Delete question (ask if entities involve user data):
4. Use AskUserQuestion:

• Question: "What delete strategy for records?"
• Header: "Deletes"
• Options:
  • "Soft delete via deleted_at (Recommended)" — Preserves data, supports undelete
  • "Hard delete" — Permanent removal, simpler queries
  • "Archival" — Move to cold storage tables
  • "Mixed per entity" — Different strategy per entity type

Retention question (ask if user data or compliance mentioned):
5. Use AskUserQuestion:

• Question: "Are there data retention or GDPR requirements?"
• Header: "Retention"
• Options:
  • "No requirements" — Standard data lifecycle
  • "GDPR compliance needed" — Right-to-deletion, crypto-shredding options
  • "TTL-based cleanup" — Automatic expiration of old records
  • "Custom retention policy" — Per-entity retention rules

Depth question (always ask):
6. Use AskUserQuestion:

• Question: "What level of detail should the output include?"
• Header: "Depth"
• Options:
  • "Production (Recommended)" — Full tables, indexes, migrations, retention, quality scorecard
  • "MVP" — Core tables and constraints only, minimal migration notes
  • "Schema only" — Tables and relationships, no migration or retention notes

Step 3: Entity-Relationship Analysis

For each entity, apply constraint extraction heuristics:

Constraint Extraction Rules

Reference: See ${CLAUDE_PLUGIN_ROOT}/docs/extending/backend-data-model-reference.md section "Constraint Extraction Patterns" for uniqueness detection, relationship mapping, CHECK constraint patterns, and NOT NULL defaults.

Critical multi-tenant rule: When multi-tenancy is enabled, every uniqueness constraint must include tenant_id — UNIQUE(tenant_id, email), never UNIQUE(email). This is the most common AI failure in schema generation.

Per-Entity Analysis

For each entity, determine:

Attribute	Detail
Table name	Plural snake_case (e.g., order_items)
PK	bigint (GENERATED ALWAYS AS IDENTITY / AUTO_INCREMENT) or uuid
Columns	Name, type (engine-specific), nullable, default, constraints
Relationships	Cardinality, FK column, ON DELETE behavior
Indexes	Apply ESR rule for composites (Equality → Sort → Range)
Constraints	UNIQUE, CHECK, NOT NULL, FK

ESR Composite Index Rule

For composite indexes, always order columns:

1. Equality columns first (=, IN)
2. Sort columns next (ORDER BY, matching direction)
3. Range columns last (>, <, BETWEEN)

Example: WHERE tenant_id = ? AND status = 'active' AND created_at > ? → Index: (tenant_id, status, created_at)

Present entity map:

ENTITY MAP
──────────
Entity: User
  Table:      users
  PK:         id (bigint)
  Columns:    email (varchar, NOT NULL, UNIQUE), name (varchar, NOT NULL), role (varchar, CHECK), bio (text, nullable), created_at, updated_at
  Relations:  1:N → posts, 1:N → comments
  Indexes:    (email) UNIQUE, (created_at)
  Migration:  Greenfield — CREATE TABLE

Entity: Post
  Table:      posts
  PK:         id (bigint)
  Columns:    title (varchar, NOT NULL), body (text, NOT NULL), status (varchar, CHECK: draft/published), user_id (bigint, FK, NOT NULL), created_at, updated_at
  Relations:  N:1 → users, 1:N → comments
  Indexes:    (user_id), (status, created_at) — ESR: equality then range
  Migration:  Greenfield — CREATE TABLE

Step 4: Cross-Cutting Concerns

Plan cross-cutting patterns based on Step 2 decisions:

Reference: See ${CLAUDE_PLUGIN_ROOT}/docs/extending/backend-data-model-reference.md section "Cross-Cutting Concern Patterns" for timestamps, soft deletes, multi-tenancy (incl. RLS template), enum strategy, PK strategy, and naming conventions.

---

HARD STOP — Review Data Model Plan

Present the complete analysis summary:

DATA MODEL PLAN
═══════════════

SUMMARY
───────
Engine:        {from tech.md or Step 2}
Migration:     {Greenfield/Brownfield/Mixed}
Tenancy:       {None/Shared+tenant_id/Schema-per-tenant/DB-per-tenant}
Deletes:       {Soft/Hard/Archival/Mixed}
Retention:     {None/GDPR/TTL/Custom}
Depth:         {Production/MVP/Schema}

ENTITIES ({count})
──────────────────
| Entity | Table | Columns | Relationships | Indexes | Constraints |
|--------|-------|---------|---------------|---------|-------------|
| User | users | 6 | 1:N posts, 1:N comments | 3 | 2 UNIQUE, 1 CHECK |
| Post | posts | 7 | N:1 users, 1:N comments | 3 | 1 CHECK |
| ... | ... | ... | ... | ... | ... |

CROSS-CUTTING
─────────────
Timestamps:    created_at + updated_at on all tables
Soft Deletes:  {enabled/disabled} {+ partial unique indexes if enabled}
Multi-Tenancy: {strategy + tenant_id placement}
Enum Strategy: CHECK on VARCHAR (no native ENUM)
PK Strategy:   {bigint/uuid}
Naming:        plural snake_case tables, GitLab-style constraint naming

OUTPUT
──────
Folder: $JAAN_OUTPUTS_DIR/backend/data-model/{id}-{slug}/
File:   {id}-{slug}.md

Use AskUserQuestion:

• Question: "Proceed with generating the data model document?"
• Header: "Generate"
• Options:
  • "Yes" — Generate the data model
  • "No" — Cancel
  • "Edit" — Let me revise the scope or design first

Do NOT proceed to Phase 2 without explicit approval.

---

PHASE 2: Generation (Write Phase)

Step 5: Generate Data Model Document

Read template: $JAAN_TEMPLATES_DIR/jaan-to:backend-data-model.template.md

If tech stack needed, extract sections from tech.md:

• Current Stack: #current-stack
• Constraints: #constraints
• Patterns: #patterns

Generate the document in this order:

5.1: Executive Summary

1-2 sentences describing: entity count, database engine, key design decisions (tenancy, delete strategy), and migration approach.

5.2: Entity-Relationship Diagram

Generate Mermaid erDiagram with all entities, relationships, and cardinality.

5.3: Table Definitions

For each entity, generate:

Column table:

Column	Type	Nullable	Default	Constraints
id	BIGINT GENERATED ALWAYS AS IDENTITY	NO	—	PRIMARY KEY
email	VARCHAR(255)	NO	—	UNIQUE, NOT NULL
status	VARCHAR(20)	NO	'pending'	CHECK (status IN (...))
user_id	BIGINT	NO	—	FK → users.id ON DELETE CASCADE
created_at	TIMESTAMPTZ	NO	now()	—
updated_at	TIMESTAMPTZ	NO	now()	—

Reference: See ${CLAUDE_PLUGIN_ROOT}/docs/extending/backend-data-model-reference.md section "Engine-Specific Type Rules" for PK, timestamp, JSON, and boolean type mappings per engine.

Indexes table:

Name	Columns	Type	Rationale
idx_posts_on_user_id	(user_id)	B-tree	FK lookup performance
idx_posts_on_status_created	(status, created_at)	B-tree	ESR: equality then range

Foreign Keys table:

Column	References	ON DELETE	ON UPDATE
user_id	users.id	CASCADE	CASCADE

Migration Notes (per table):

• Greenfield: CREATE TABLE statement
• Brownfield: Zero-downtime steps using engine-appropriate patterns:
  • PostgreSQL: CREATE INDEX CONCURRENTLY, NOT VALID + VALIDATE CONSTRAINT, SET lock_timeout
  • MySQL: Check INSTANT/INPLACE/COPY algorithm; use pt-osc or gh-ost for COPY operations
  • Expand-contract pattern for column renames, type changes, structural refactoring

5.4: Cross-Cutting Concerns

Document the patterns chosen in Step 4 with concrete implementation details.

5.5: Index Strategy

For each composite index, show ESR rationale.

Reference: See ${CLAUDE_PLUGIN_ROOT}/docs/extending/backend-data-model-reference.md section "Index Strategy Patterns" for engine-specific index types, multi-tenant indexing, partial indexes, and covering index patterns.

5.6: Migration Playbook

Per-table migration classification.

Reference: See ${CLAUDE_PLUGIN_ROOT}/docs/extending/backend-data-model-reference.md section "Migration Safety Classification" for operation safety levels, methods, and brownfield zero-downtime patterns.

5.7: Retention & Compliance

If GDPR: document deletion strategy (hard delete + anonymized audit, crypto-shredding, or separate PII tables).
If TTL: document cleanup approach (pg_cron + batched DELETE, partition-based retention, MongoDB TTL indexes, DynamoDB TTL).
If legal holds: note legal_holds table pattern.

5.8: Quality Scorecard

Apply 5-dimension scoring rubric. Score each dimension and compute weighted overall score.

Reference: See ${CLAUDE_PLUGIN_ROOT}/docs/extending/backend-data-model-reference.md section "Quality Scorecard Rubric" for dimensions, weights, and check criteria.

Step 6: Quality Check

Before preview, verify every item in the quality checklist. If any check fails, fix before preview.

Reference: See ${CLAUDE_PLUGIN_ROOT}/docs/extending/backend-data-model-reference.md section "Quality Check Checklist" for structure, constraints, indexes, anti-patterns, and completeness checks.

Step 7: Preview & Approval

Show the complete data model document.

Use AskUserQuestion:

• Question: "Write the data model document to output?"
• Header: "Write"
• Options:
  • "Yes" — Write the file
  • "No" — Cancel
  • "Edit" — Let me revise something first

Step 7.5: Generate ID and Folder Structure

If approved, set up the output structure:

1. Source ID generator utility:

source "${CLAUDE_PLUGIN_ROOT}/scripts/lib/id-generator.sh"

2. Generate sequential ID and output paths:

# Define subdomain directory
SUBDOMAIN_DIR="$JAAN_OUTPUTS_DIR/backend/data-model"
mkdir -p "$SUBDOMAIN_DIR"

# Generate next ID
NEXT_ID=$(generate_next_id "$SUBDOMAIN_DIR")

# Create folder and file paths (slug from entity/feature name)
slug="{lowercase-hyphenated-name-max-50-chars}"
OUTPUT_FOLDER="${SUBDOMAIN_DIR}/${NEXT_ID}-${slug}"
MAIN_FILE="${OUTPUT_FOLDER}/${NEXT_ID}-${slug}.md"

3. Preview output configuration:

Output Configuration

• ID: {NEXT_ID}
• Folder: $JAAN_OUTPUTS_DIR/backend/data-model/{NEXT_ID}-{slug}/
• Main: {NEXT_ID}-{slug}.md

Step 8: Write Output

1. Create output folder:

mkdir -p "$OUTPUT_FOLDER"

2. Write data model document to main file.

3. Update subdomain index:

source "${CLAUDE_PLUGIN_ROOT}/scripts/lib/index-updater.sh"
add_to_index \
  "$SUBDOMAIN_DIR/README.md" \
  "$NEXT_ID" \
  "${NEXT_ID}-${slug}" \
  "{Data Model Title}" \
  "{1-2 sentence executive summary}"

4. Confirm completion:

✓ Data model written to: $JAAN_OUTPUTS_DIR/backend/data-model/{NEXT_ID}-{slug}/{NEXT_ID}-{slug}.md
✓ Index updated: $JAAN_OUTPUTS_DIR/backend/data-model/README.md

Step 9: Suggest Next Steps

"Data model generated. Suggested next steps:"

1. API contract: Generate OpenAPI spec from this data model:

   /jaan-to:backend-api-contract "{entity-list}"

2. Task breakdown: Generate backend tasks from this data model:

   /jaan-to:backend-task-breakdown "{prd-or-feature}"

3. Review: Have the team review constraint completeness and index strategy

Step 10: Capture Feedback

Use AskUserQuestion:

• Question: "Any feedback on the generated data model?"

• Header: "Feedback"

• Options:

  • "No" — All good, done
  • "Fix now" — Update something in the data model
  • "Learn" — Save lesson for future runs
  • "Both" — Fix now AND save lesson

• Fix now: Update the output file, re-preview, re-write

• Learn: Run /jaan-to:learn-add backend-data-model "{feedback}"

• Both: Do both

---

Skill Alignment

• Two-phase workflow with HARD STOP for human approval
• Multi-stack support via tech.md detection
• Template-driven output structure
• Output to standardized $JAAN_OUTPUTS_DIR path

Definition of Done

• Input parsed, entities identified and confirmed
• Data design decisions confirmed (engine, migration, tenancy, deletes, retention, depth)
• Entity relationships mapped with columns, constraints, indexes
• Cross-cutting concerns documented (timestamps, soft deletes, tenancy, audit, enums)
• Data model document generated with Mermaid ER diagram
• Migration notes included per table (zero-downtime for brownfield)
• Quality checks passed (anti-patterns + 5-dimension rubric)
• Output written to $JAAN_OUTPUTS_DIR/backend/data-model/{id}-{slug}/
• Subdomain index updated
• User approved final result