setup-datamodel

📋 Shared instructions: shared-instructions.md — read first.

Set Up Data Model + Connectors

Combined orchestrator for standalone data source planning. Designs the Dataverse schema, plans connectors, gets approval on both, then delegates execution to /add-dataverse and /add-connector.

Workflow

1. Verify project & auth → 2. Design data model → 3. Plan connectors → 4. Combined approval → 5. Execute data model → 6. Execute connectors → 7. Summary

Phase 1 — Verify Project & Auth

Confirm we're inside a Power Apps mobile app:

test -f power.config.json && echo "OK" || echo "ERROR: not a mobile app — run /create-mobile-app first"
node "${CLAUDE_SKILL_DIR}/../../scripts/resolve-environment.js" "$(node -e \"console.log(require('./power.config.json').environmentId)\")"

Capture the environment URL, environment ID, tenant ID, and organization ID for Phase 5.

Phase 2 — Design Data Model

Check $ARGUMENTS for diagram hints first (*.png, *.jpg, erDiagram keyword, ||--o{ cardinality syntax). If a hint is present → Path A. If $ARGUMENTS describes the app at all → silently take Path B (architect propose). Only if both are empty, ask:

"How would you like to define the data model?"

Default the answer to "architect propose" so an empty answer auto-proceeds without blocking the user.

Artifact storage rules for PDFs and signatures

When requirements mention signatures, sign-off, ink, drawings, generated PDFs, exported reports, evidence packets, or retained documents, make the storage target explicit in ## Data Model before approval:

PDF content must never be modeled as long text/base64 text. Use Dataverse File columns for retained PDFs. Signature PNGs may use Image columns when the generated service supports image payloads; use File columns or child Evidence rows when the capture should behave like an attachment.

Path A — Parse user-provided diagram

Accept PNG/JPG (use Read to view), Mermaid syntax (paste in chat), or text description. Parse into tables + columns + relationships. Query existing Dataverse tables to mark each as new / extend / reuse. Generate a Mermaid ER diagram for confirmation. Enter EnterPlanMode for data model approval. On ExitPlanMode approval, write the data model into native-app-plan.md ## Data Model section (creating the file if absent).

Path B — Spawn data-model-architect

Task: mobile-app:data-model-architect

Prompt:
  You are the data-model-architect agent for a Power Apps mobile app.
  Requirements: <$ARGUMENTS or ask the user what the app does>
  Working directory: <cwd>
  Plugin root: ${CLAUDE_SKILL_DIR}/../../

  Follow your agent file. Return a ## Data Model section with Mermaid ER diagram,
  reuse/extend/create table, and dependency-tier ordering. If requirements mention
  signatures, pen/ink, generated PDFs, report exports, evidence packets, or uploaded
  documents, include the artifact storage target: on-device/share-only, Dataverse
  Image column, Dataverse File column, or child Evidence/Attachment table. Retained
  PDF content must use a File column, not long text/base64.

Present the returned section via EnterPlanMode / ExitPlanMode for approval.

Path C — No Dataverse

Write ## Data Model as "None — no Dataverse tables needed." Continue to Phase 3.

Phase 3 — Plan Connectors

Follow shared/references/connector-planning.md:

1. Infer — if $ARGUMENTS describes what the app does, scan for connector keywords. Build a candidate list.
2. Confirm — present via AskUserQuestion. Let the user add, remove, or confirm.
3. Record — build the ## Connectors section.

If the user provided no requirements context, ask:

"What does your app need to connect to? (e.g. SharePoint, Teams, email, Excel, OneDrive, Azure DevOps — or none)"

Phase 4 — Combined Approval

Present the full plan — data model + connectors — together in a single EnterPlanMode block:

## Plan: Data Sources

### Data Model
[reuse/extend/create table]
[Mermaid ER diagram]
[creation order tiers]

### Connectors
[connector table or "None"]

Approve both to proceed with execution?

• Approved → proceed to Phase 5
• Change data model → loop back to Phase 2 for that section only, then re-present Phase 4
• Change connectors → loop back to Phase 3, then re-present Phase 4

Phase 5 — Execute Data Model

Invoke /add-dataverse with --skip-planning so it reads the approved plan directly without re-prompting:

Invoke skill: /add-dataverse

Arguments:
  --working-dir <cwd>
  --plan-section native-app-plan.md#data-model
  --skip-planning

/add-dataverse creates tables in tier order, runs npx power-apps add-data-source --api-id dataverse --org-url <envUrl> --resource-name <name> per table from the app root, publishes customizations, writes .datamodel-manifest.json, and type-checks. Wait for it to return before Phase 6.

Calculated columns from the screen plan — if ## Data Model in the plan includes a ### Cross-entity Reads (auto-derived from screen plan) subsection (the data-model-architect's Step 6a addendum, approved at the Gate 1 addendum during planning), /add-dataverse Step 5c creates each row as a calculated column on the parent table via scripts/create-calculated-column.js. No additional action needed in this skill — /add-dataverse handles it. If the subsection is absent, Step 5c is silently skipped. See shared/references/data-performance.md § Cross-entity Reads for why this matters.

Skip if Phase 2 chose Path C (no Dataverse).

Phase 6 — Execute Connectors

Read ## Connectors from native-app-plan.md. For each connector row, invoke /add-connector:

Invoke skill: /add-connector

Arguments:
  --working-dir <cwd>
  --connector <api-name>

Run sequentially. Skip if ## Connectors is "None".

Phase 7 — Summary

✅ Data sources set up
─────────────────────────────────────────────
Data model:
  Tables reused  : <list>
  Tables extended: <list>
  Tables created : <list>
  Manifest       : .datamodel-manifest.json

Connectors:
  <list of added connectors, or "None">

Generated services:
  src/generated/services/ × <N>
  src/generated/models/   × <N>

Type-check: PASS

Next steps:
  /add-datasource   — add more data sources
  /add-native       — add device capabilities
  screen-builder    — implement screens using the generated services
─────────────────────────────────────────────

Reference

• shared/references/connector-planning.md — connector inference + confirmation logic
• skills/add-dataverse/SKILL.md — full data model execution workflow
• agents/data-model-architect.md — read-only architect agent