data_playground

Explain a metric spike or dip by slicing dimensions and checking upstream changes. Produces a hypothesis with evidence; never auto-fixes the data. Use when a tracked metric is out of expected range but no pipeline error or quality check fired.

Find Bruin assets past their inferred freshness expectations and classify why. Distinguishes between "upstream is late", "source is down", "scheduler missed a tick", "table stopped advancing", and "this is a real outage". Use on a recurring schedule or when triage flags 'stale'.

Open a pull request for routine pipeline maintenance proposed by another skill (column rename, type bump, removed-column cleanup, dependency update, check threshold adjustment). Never opens PRs unprompted; always gated on a prior skill's finding.

Safely trigger or rerun Bruin Cloud pipeline runs for a specific time range. Validates scope, checks dependencies and materialization risk, and executes only through Bruin Cloud MCP or Bruin Cloud CLI. Use when a fix has been merged, a transient failure needs a retry, or historical data was wrong and needs regeneration.

Deep-dive on a single failed or suspicious Bruin asset. Pulls logs, error patterns, dependency graph, and recent commits to produce a root-cause hypothesis. Use when triage classified an issue as 'unknown', or a human asks "why did X fail".

Entry point for self-healing pipeline work. Pull current pipeline state from Bruin Cloud, classify what is wrong, and route to the right follow-up skill. Use this when an alert fires, a scheduled tick runs, or a human asks "what is broken right now".

Investigate a failed Bruin quality check. Walk the failing rows back to their source, classify the failure mode, and either propose a fix or hand off. Use when a custom check or column check fails on an otherwise successful run.

Detect and respond to schema drift between a source system and a Bruin asset's declared columns. Compares live schema to the asset definition, classifies the drift, and either opens a maintenance PR or escalates. Use when diagnose identifies a column/type mismatch error.

Create DAC dashboards by writing YAML or TSX dashboard definition files. Use when the user wants to create, modify, review, or understand DAC dashboards, widgets, filters, SQL queries, semantic models, or CLI validation workflows.