AGENT SKILL

analytics-engineering-architect

Architect-level analytics engineering — dbt project structure, staging and mart conventions, metric definitions and the semantic layer, incremental models, testing and documentation as code, and the review process that keeps SQL maintainable. Use when analysts and engineers share a SQL transformation layer.

License 许可
MIT
最近更新
2026-11-13
Stars
★ 0
Forks
0

01 前言 Front Matter

name
analytics-engineering-architect
license
MIT
metadata
{ "crusor-path": "data-engineering-architect/transform-architect/analytics-engineering-architect", "crusor-domain": "data-engineering-architect", "crusor-origin": "original", "crusor-sources": "docs.getdbt.com; docs.getdbt.com/docs/build/semantic-models; cube.dev/docs; sqlmesh.readthedocs.io", "crusor-triggers": "dbt, analytics engineering, semantic layer, metrics, sql models, staging models, lookml, metric definition, data mart, transformed model, disagree, revenue number, metric drift", "crusor-verified-on": "2026-10-30" }
description
Architect-level analytics engineering — dbt project structure, staging and mart conventions, metric definitions and the semantic layer, incremental models, testing and documentation as code, and the review process that keeps SQL maintainable. Use when analysts and engineers share a SQL transformation layer.
compatibility
dbt Core/Cloud, SQLMesh, LookML, Cube, MetricFlow

03 SKILL.md 正文

Analytics Engineering Architect

The transformation layer is software with consumers. It needs modules, tests, reviews, and one definition per metric — or it becomes a folder of queries with tense relationships.

When to use

  • Analysts and engineers share SQL transformations.
  • Building or untangling a dbt project.
  • Two dashboards disagree on a number.

When NOT to use

  • One-off analysis in a notebook.
  • Pipeline extraction and loading → batch-pipeline-architect.

Decision 1 — Project structure

models/
  staging/     stg_<source>__<entity>     one per source table, renamed & typed
  intermediate/ int_<entity>__<verb>      joins and business logic, not exposed
  marts/       fct_/dim_/obt_<domain>     consumed by BI, owned by a domain
  • Staging does only rename, cast, deduplicate. No joins, no business logic.
  • Intermediate models keep marts readable; they are private by default.
  • Marts are the public API: named, documented, tested, and owned.
  • One source definition per source table; freshness checked at the source.

Decision 2 — Materialisation

Model Materialisation
staging view
intermediate ephemeral or view
dimension table
fact (large, incremental) incremental with a merge key
mart for BI table or view depending on query cost
  • Views are free to build and expensive to query; tables are the reverse.
  • Incremental only with a correct unique key and a late-data strategy.
  • Iceberg/Delta dynamic tables where the warehouse supports them — less bespoke incremental logic.

Decision 3 — Metrics

  • A metric is defined once: name, description, formula, grain, filters, owner.
  • Encode it in the semantic layer (MetricFlow/Cube/LookML), not in 12 dashboards.
  • Define the entity graph so a metric can be sliced consistently across facts.
  • Changing a metric is a versioned event with an announcement, not a quiet edit.
  • Track metric lineage to source tables; a metric that cannot explain itself is a rumour.

Decision 4 — Testing and docs as code

  • Every mart: uniqueness and not-null on its key, relationships to dimensions, accepted values for enums, and a freshness check.
  • Document every model with purpose, grain, and owner — generated docs, not a wiki.
  • Exposures declare dashboards and downstream consumers; that powers impact analysis.
  • CI: dbt build on a sample schema + linting + contract tests before merge.

Decision 5 — Review and ownership

  • SQL gets code review like any code: readability, joins that can fan out, filters that silently drop rows.
  • Domain ownership: each mart has a named team responsible for its correctness.
  • Deprecation: mark, announce with a date, then remove. Never delete a used model on a Friday.
  • Cost review: which models are rebuilt hourly and read twice a day?

Failure modes at scale

  • 400 models with no staging layer; every mart re-implements the same casts.
  • Three definitions of revenue, all "official", in three tools.
  • Incremental model with no unique key, slowly accumulating duplicates.
  • Docs generated once at launch, stale within a month.
  • A mart deleted by a cleanup pass; four dashboards break the next morning.

Verification receipts

  1. Layer conventions documented and enforced in CI/lint.
  2. Metric definitions register with one owner each; no duplicates.
  3. dbt build green on a CI sample including tests and contracts.
  4. Fan-out test: fact-to-dimension joins produce no row multiplication.
  5. Deprecation log with consumers mapped via exposures.

04 同仓库更多 Skills

SriAkhilSJ
SriAkhilSJ/PulseAIRepo

interaction-patterns-architect

Architect-level interaction design in code — choosing established patterns over invented ones, feedback and latency behaviour, destructive-action safeguards, navigation and overlay conventions, and the state model behind every interactive surface. Use when a screen has to behave, not just look.

★ 0
SriAkhilSJ
SriAkhilSJ/PulseAIRepo

design-system-ops-architect

Architect-level operation of a design system in code — token pipelines and versioning, component API and variant policy, the catalogue as the contract, adoption metrics, and the governance that stops a system from forking into three. Use when a component library must survive several teams.

★ 0
SriAkhilSJ
SriAkhilSJ/PulseAIRepo

design-handoff-architect

Architect-level design handoff — what must be specified before implementation starts, tokens over raw values, responsive and content-limit contracts, state matrices, and the review loop that catches drift between design and code. Use when a design is about to become code.

★ 0
SriAkhilSJ
SriAkhilSJ/PulseAIRepo

content-microcopy-architect

Architect-level UI content — where strings live, voice and terminology consistency, error and empty-state copy as part of the interface, translation-ready authoring, and the content model a product needs before it multiplies screens. Use when the words in the interface matter as much as the pixels.

★ 0
SriAkhilSJ
SriAkhilSJ/PulseAIRepo

ui-ux-implementation-architect

Architect-level direction for turning a design into working UI — the handoff contract, component ownership versus screen-level layout, interaction and state conventions, design QA gates, and the review habits that keep an implementation faithful without frozen pixels. Use before any specific UI implementation skill.

★ 0
SriAkhilSJ
SriAkhilSJ/PulseAIRepo

test-strategy-architect

Architect-level test strategy — allocating assertions across unit, integration, contract and end-to-end layers, designing for testability through dependency boundaries, fixture and data strategy, and the flake and coverage policies that keep a suite trustworthy. Use when defining how a codebase is tested.

★ 0