nledford/engineering-review-board / skills/data-platform-engineering
data-platform-engineering
VerifiedData-platform engineering guidance for implementing, changing, reviewing, or testing source-to-landing ingestion, post-landing lakehouse or warehouse transformations, analytical semantic contracts, Power BI semantic models and DAX, and Microsoft Fabric or Power BI operations. Use for CDC, watermarks, replay, Delta tables, medallion layers, analytical grain, governed metrics, TMDL, RLS or OLS, refresh, promotion, gateways, capacity, recovery, and runbooks. Do not use for application domain modeling, physical OLTP database design, generic distributed protocols, generic UX, or final release decisions.
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 11.7 KB |
| references/analytical-semantics.md | markdown | 6.0 KB |
| references/analytics-transformations.md | markdown | 5.0 KB |
| references/fabric-power-bi-operations.md | markdown | 6.0 KB |
| references/ingestion.md | markdown | 5.8 KB |
| references/power-bi-semantic-models.md | markdown | 5.9 KB |
| references/pyspark-notebooks.md | markdown | 20.6 KB |
references/power-bi-semantic-models.md
View on GitHub# Power BI Semantic Models
Load this reference when Power BI model behavior is material: model shape,
relationships, DAX, storage mode, refresh semantics, RLS or OLS, report-facing
compatibility, or query performance. Load
[analytical semantics](analytical-semantics.md) too when governed meaning, grain,
identity, history, or metric ownership changes.
## Inspect the Model as Source
- Prefer repository-owned project, TMDL, model, deployment, test, and sample-query
artifacts over screenshots or manually transcribed UI state.
- Identify generated files and their source of truth before editing. Keep changes
narrow enough that model diffs remain reviewable.
- Inventory tables, columns, measures, hierarchies, relationships, calculation
groups, roles, perspectives, partitions, storage modes, refresh policies, and
consumers affected by the change.
- Treat undocumented tenant settings, service behavior, feature availability,
licensing, and defaults as unknown until verified against current authoritative
evidence.
## Model Shape and Relationships
- State the grain, key, uniqueness, null behavior, and role of every affected
table. Prefer a clear fact-and-dimension shape when it matches the domain, but
do not force one when a different shape has an evidenced purpose.
- Validate relationship endpoints, cardinality, active state, referential
assumptions, and filter direction against actual key profiles.
- Use bidirectional filtering and many-to-many relationships only with an
explicit semantic need and tests for ambiguity, double counting, and query
behavior.
- Model role-playing dates, inactive relationships, bridge tables, unknown
members, and slowly changing dimensions intentionally.
- Keep technical columns hidden when they do not help consumers; expose stable
names, descriptions, formats, folders, sort-by behavior, hierarchies, and
default summarization that support correct self-service use.
## DAX Correctness
- Write the metric contract before the expression: base grain, formula, filters,
inclusion rules, time basis, dimensional validity, blank or zero behavior, and
expected totals.
- Reason explicitly about filter context, row context, context transition,
expanded tables, relationship propagation, and evaluation at detail, subtotal,
and grand-total levels.
- Test single and multi-select filters, no selection, missing keys, blanks, zero
denominators, negative values, duplicate identifiers, many-to-many paths, and
non-additive totals where relevant.
- Prefer reusable base measures and explicit compositions over duplicated report-
local logic. Remove unused objects only after checking dependencies and
compatibility.
- Treat time-intelligence behavior as a calendar and data-completeness contract,
not a function-name shortcut.
## Storage, Refresh, and Compatibility
- Choose Import, DirectQuery, Direct Lake, composite, aggregation, or hybrid
behavior from freshness, latency, volume, concurrency, source load, security,
cost, and recovery evidence. Platform support and limitations are
version-sensitive.
- Define partition boundaries, incremental-refresh or change-detection semantics,
late data, correction, deletion, restatement, and full-refresh fallback.
- Verify model changes against representative existing reports, queries,
bookmarks, field references, exports, APIs, and downstream models.
- Classify renames, removals, type changes, relationship changes, format changes,
and metric-definition changes by compatibility impact; provide a migration or
versioning path for breaking changes.
## Model Security
- Treat RLS and OLS as authorization controls. Define identity mapping, role
membership, tenant or organizational scope, default-deny behavior, and how
unknown or malformed identities behave.
- Test authorized and unauthorized personas, multiple-role combinations,
relationship propagation, bridge behavior, hidden versus protected objects,
export paths, and any privileged bypass.
- Keep credentials, service principals, gateway secrets, and sensitive sample
values outside model source, logs, screenshots, and test reports.
- Verify that upstream permissions, model permissions, build access, sharing,
downstream models, and report delivery do not undermine the intended policy.
## Performance and Usability Evidence
- Use representative report queries, filter combinations, concurrency, data
volume, refresh windows, and cold or warm cache conditions. Record the tools,
environment, and limits of the measurement.
- Inspect expensive expressions, high-cardinality columns, relationship paths,
storage-engine versus formula-engine work, partition pruning, source queries,
and refresh resource use when supported by available evidence.
- Do not trade correctness or maintainability for a micro-optimization without a
measured bottleneck and regression coverage.
- Check discoverability, naming, descriptions, formats, folders, default behavior,
error or blank states, and localization when the model is a self-service
interface. Route visual-layout and accessibility findings to UX review.
## Validation Evidence
- Validate model metadata or source with repository-supported tooling.
- Run compact DAX query fixtures for expected detail rows, filter combinations,
subtotals, totals, blanks, edge dates, security personas, and compatibility.
- Reconcile representative model results to the governed metric contract and
published analytical tables; explain every material variance.
- Exercise refresh success, empty increment, late correction, interruption,
retry, and full fallback when refresh behavior changes.
- Capture representative query and refresh performance before and after material
performance changes.
Before handoff, report the affected semantic contract, model objects, security
behavior, refresh and compatibility impact, executable evidence, version-sensitive
assumptions, and remaining report or operational risk.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| finishing-a-development-branchobra/superpowersUse when implementation is complete, all tests pass, and you need to decide how to integrate the work | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| systematic-debuggingobra/superpowersUse when encountering any bug, test failure, or unexpected behavior, before proposing fixes | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| test-driven-developmentobra/superpowersUse when implementing any feature or bugfix, before writing implementation code | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| skill-creatoranthropics/skillsCreate new skills, modify and improve existing skills, and measure skill performance. Use when users want to create a skill from scratch, edit, or optimize an existing skill, run evals to test a skill, benchmark skill performance with variance analysis, or optimize a skill's description for better triggering accuracy. | Stars: 198k | Updated | Claude CodeCodex | Verified | Install |