菜单
Droughty schema-introspection toolkit — use when working with Droughty commands, profile/project configuration, LookML generation from warehouse schemas, dbt test generation, DBML diagrams, field documentation, or data quality reports
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
| name | droughty |
| description | Droughty schema-introspection toolkit — use when working with Droughty commands, profile/project configuration, LookML generation from warehouse schemas, dbt test generation, DBML diagrams, field documentation, or data quality reports |
| tags | ["droughty","lookml","dbt","schema","bigquery","snowflake","dbml","qa"] |
This skill activates when working on Droughty-related tasks within a Wire engagement. It provides guidance on the Droughty CLI, configuration, Wire integration conventions, and output structure.
Droughty is a schema-driven toolkit that reads the live warehouse and generates semantic layer artifacts. It is the bottom-up counterpart to Wire's top-down, document-driven workflow.
| Command | What it does | Requires |
|---|---|---|
droughty dbml | DBML entity-relationship diagram | Warehouse connection |
droughty docs | AI field descriptions for all columns | Warehouse + OpenAI API key |
droughty qa | LangGraph data quality agent | Warehouse + OpenAI API key |
droughty dbt | Pattern-based schema.yml tests | Deployed dbt tables |
droughty stage | Staging SQL + sources.yml (BigQuery only) | Source dataset in BigQuery |
droughty lookml | Base LookML views, explores, measures | Deployed dbt tables |
Wire pins a specific Droughty version in wire/droughty/pinned_version.txt. All /wire:droughty-setup runs install this version. Current pin: check the file at wire/droughty/pinned_version.txt.
To update the pin, run the refresh script (Wire repo owner only):
bash wire/droughty/refresh_version.sh # updates pinned_version.txt
bash wire/droughty/refresh_version.sh --commit # also commits the change
Consultants re-run /wire:droughty-setup --force after pulling the updated repo.
~/.droughty/profile.yaml)Not committed to git — contains warehouse credentials.
BigQuery:
my_engagement:
type: bigquery
project: my-gcp-project
dataset: analytics
schemas:
- analytics
- staging
openai_api_key: sk-... # required for docs + qa
Snowflake:
my_engagement:
type: snowflake
account: xy12345.us-east-1
username: analyst
password: secret
warehouse: COMPUTE_WH
database: ANALYTICS
schema: PUBLIC
role: ANALYST
schemas:
- PUBLIC
- STAGING
openai_api_key: sk-...
Committed to git — controls output paths.
profile_name: my_engagement
# Wire-aligned output paths
dbml_path: .wire/releases/01-discovery/artifacts/droughty/
field_description_path: .wire/releases/01-discovery/artifacts/droughty/field_descriptions/
dbt_path: ./models/
stage_path: ./models/staging/
lookml_path: ./lookml/views/generated/
Droughty-generated views land in views/generated/. Wire-authored extensions (explores, refinements, derived fields) go in views/extended/. Never edit views/generated/ by hand — Droughty regenerates it on each /wire:droughty-lookml run.
lookml/
├── views/
│ ├── generated/ ← Droughty output — auto-regenerated, do not hand-edit
│ │ ├── orders.view.lkml
│ │ └── customers.view.lkml
│ └── extended/ ← Wire extensions — business logic, explores, refinements
│ ├── orders_extended.view.lkml
│ └── explores.model.lkml
Use LookML refinements in views/extended/ to extend base views:
view: +orders {
dimension: order_value_band {
type: string
sql: CASE WHEN ${order_total} < 100 THEN 'low' ...
}
}
| Command | Purpose |
|---|---|
/wire:droughty-setup <release> | Install Droughty, generate profile.yaml + droughty_project.yaml |
/wire:droughty-introspect <release> | Schema inventory report — tables, columns, PK/FK coverage |
/wire:droughty-dbml <release> | DBML entity-relationship diagram |
/wire:droughty-docs <release> | AI field descriptions |
/wire:droughty-qa <release> | Data quality agent report |
/wire:droughty-stage <release> | Staging SQL + sources.yml (BigQuery only) |
/wire:droughty-dbt-tests <release> | Pattern-based schema tests (post-dbt deploy) |
/wire:droughty-lookml <release> | Base LookML views (post-dbt deploy) |
/wire:droughty-generate <release> | Full Droughty phase in sequence |
Discovery / existing warehouse audit (Droughty release type):
/wire:new (release_type: droughty)
/wire:droughty-setup
/wire:droughty-introspect
/wire:droughty-dbml
/wire:droughty-docs
/wire:droughty-qa
→ feed artifacts into /wire:problem-definition-generate
Post-dbt phase (within full_platform or dbt_development):
... /wire:dbt-validate ← dbt run ← /wire:dbt-generate ...
/wire:droughty-setup
/wire:droughty-dbt-tests
/wire:droughty-lookml
/wire:droughty-docs
/wire:droughty-qa
/wire:semantic_layer-generate ← extends Droughty base views
droughty: command not found — install with pip install "droughty==[version]". Python 3.9–3.12.3 required.
profile.yaml not found — run /wire:droughty-setup [release].
No tables found — check that schemas: in profile.yaml matches the actual schema names in the warehouse. BigQuery schema names are case-sensitive.
OpenAI API error — check that openai_api_key is set in profile.yaml and the key has billing enabled.
droughty qa timeout — the QA agent runs live warehouse queries; large schemas can take 20+ minutes. Consider limiting schemas in scope.
Generates a delivery-focused status report for a named Rittman Analytics client engagement. Anchored on SOW milestones and sprint cadence — not a relationship or commercial report. Reports milestone status (on track / at risk / slipping), tickets and MRs delivered, bugs raised and resolved, decisions owed by the client, cross-cutting risks, and client and team sentiment. Pulls from Fathom meeting transcripts, client Slack channels (external and internal), and Atlassian where available. Use when Mark asks for delivery status, project status, milestone status, sprint status, "where are we on [client]", "is [client] on track", or "what's slipping on [client]". Triggers on any client name (Client A, Client B, Client C, Client D, Client E, Barton Peveril, Client F, Client H, Client I, Client J) plus a delivery framing.
Proactive skill for building Dagster orchestration layers. Auto-activates when creating or modifying Dagster assets, schedules, sensors, or components, or when working on the orchestration artifact in a Wire project. Covers the assets-first pattern, dagster-dbt integration, automation (schedules, sensors, declarative automation), the component framework, and CLI usage (dg dev, dg launch, dg check, dg scaffold).
Proactive skill for answering business analytics questions using dbt's Semantic Layer, Discovery API, or direct SQL. Auto-activates when a user asks a data or metrics question against a dbt project (e.g. "What were total sales last quarter?", "Show me top customers by revenue"). Uses a 4-level escalation — Semantic Layer first, modified compiled SQL, model discovery, manifest analysis. Always exhausts all options before saying "cannot answer."
Proactive skill for generating Mermaid flowchart diagrams of dbt model lineage. Auto-activates when a user asks to visualise dbt model dependencies, create a lineage diagram, or show the dbt DAG. Uses MCP get_lineage tools first, then manifest.json parsing, then direct code parsing as fallbacks. Produces colour-coded diagrams matching Wire's naming conventions.
Proactive skill for validating dbt models against coding conventions. Auto-activates when creating, reviewing, or refactoring dbt models in staging, integration, or warehouse layers. Validates naming, SQL structure, field conventions, testing coverage, and documentation. Supports project-specific convention overrides and sqlfluff integration.
Proactive skill for creating dbt unit tests. Auto-activates when working with dbt model testing, mock inputs/outputs, or transformation validation. Covers the Model-Inputs-Outputs pattern, format selection (dict/CSV/SQL), BigQuery-specific caveats, and production deployment guidance.