JPeetz/agent-skills / skills/data/data-pipeline-engineering
data-pipeline-engineering
VerifiedUse this skill when designing, building, or debugging data pipelines — ETL/ELT workflows, data orchestration (Airflow/Dagster/Prefect), streaming (Kafka/Spark Streaming), schema drift handling, pipeline observability, backfill strategies, and data warehouse integration (BigQuery/Snowflake/Redshift). Covers batch and real-time patterns, idempotency, partitioning, incremental loads, and monitoring.
| File | Kind | Size |
|---|---|---|
| CHANGELOG.md | markdown | 3.4 KB |
| LICENSE | other | 1.0 KB |
| SKILL.md | skill md | 26.0 KB |
| evals/evals.json | config | 5.1 KB |
| references/sources.md | markdown | 6.2 KB |
| scripts/data_freshness_check.py | script | 9.1 KB |
| scripts/pipeline_lint.py | script | 12.9 KB |
CHANGELOG.md
View on GitHub# Changelog
All notable changes to the data-pipeline-engineering agent skill.
## [1.0.0] — 2026-11-13
### Initial Release
Skill for designing, building, and debugging batch and streaming data
pipelines, with emphasis on the failure modes that actually cause production incidents
rather than tool syntax.
### What's Included
- **Core mental model** — a pipeline as a function of (source snapshot, partition, run
timestamp), and the four properties that follow from it: idempotent, partition-scoped,
observable, reversible.
- **Architecture guidance** — ETL vs ELT decision table, batch vs micro-batch vs streaming
decision table with the operational-cost caveat, and raw/staging/mart/monitoring layering.
- **Orchestration comparison** — Airflow vs Dagster vs Prefect on primitive, strengths,
weaknesses, and backfill UX, with working example DAG/asset/flow snippets and the
serialization and logical-date pitfalls called out.
- **Idempotency** — three strategies (partition overwrite, MERGE on business key, append +
dedupe view), half-open watermark windows, late/duplicate handling, full MERGE SQL, and
delete-handling options.
- **Schema drift** — fingerprinting, a drift classification table with severity and
required action, catalog-based detection per warehouse, and the
"tolerant on read, strict on write" rule.
- **Data quality** — boundary-based checks (extract / transform / publish), minimum viable
check set with blocking-vs-warning guidance, and a combined SQL check query.
- **Observability** — the four per-run signals (freshness, volume, latency, status +
schema), cost tracking, and an alert-design section (symptom-based, deduplicated,
numbered, tiered) including scheduler watchdogs.
- **Backfill and replay** — six-step procedure, idempotency proof before backfill,
throttling, atomic swap isolation, downstream cascade, and run-type tagging.
- **Warehouse-specific patterns** — BigQuery partitioning/clustering/partition-filter
enforcement, Snowflake clustering keys and warehouse separation, Redshift dist/sort keys
and VACUUM, plus cross-warehouse rules for semi-structured data, timezones, and units.
- **Anti-pattern table** — ten common mistakes mapped to symptom and fix.
- **Build checklist** and a **failing-run debugging order of operations**.
### Artifacts
- `SKILL.md` — Cross-platform skill definition
- `LICENSE` — MIT
- `CHANGELOG.md` — This file
- `evals/evals.json` — 3 evaluation scenarios (design, debug, operations)
- `references/sources.md` — Authoritative upstream documentation
- `scripts/pipeline_lint.py` — Static lint for a pipeline spec: idempotency declaration,
partition scoping, non-deterministic partition bounds, watermark/lookback, cost
guardrails, quality-check completeness, backfill bounds. Stdlib only, JSON output,
exit 1 on errors.
- `scripts/data_freshness_check.py` — Freshness lag vs SLA, missing-partition gaps (only
for windows already closed, so the in-progress period is never a false positive), and
row-count anomalies vs trailing median. Stdlib only, exit 2 on failure.
### Design Notes
- Scope deliberately excludes dbt model authoring and OLTP schema design; those route to
`dbt-data-transformation` and `database-schema-designer` respectively.
- Scripts avoid `yaml` and every third-party dependency so they can run in a bare CI
container or an agent sandbox without an install step.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| writing-skillsobra/superpowersUse when creating new skills, editing existing skills, or verifying skills work before deployment | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| xlsxanthropics/skillsUse this skill any time a spreadsheet file is the primary input or output. This means any task where the user wants to: open, read, edit, or fix an existing .xlsx, .xlsm, .xltx, .csv, or .tsv file (e.g., adding columns, computing formulas, formatting, charting, cleaning messy data); create a new spreadsheet from scratch or from other data sources; or convert between tabular file formats. Trigger especially when the user references a spreadsheet file by name or path — even casually (like "the xlsx in my downloads") — and wants something done to it or produced from it. Also trigger for cleaning or restructuring messy tabular data files (malformed rows, misplaced headers, junk data) into proper spreadsheets. The deliverable must be a spreadsheet file. Do NOT trigger when the primary deliverable is a Word document, HTML report, standalone Python script, database pipeline, or Google Sheets API integration, even if tabular data is involved. | Stars: 198k | Updated | Claude CodeCodex | Verified | Install |
| shipping-and-launchaddyosmani/agent-skillsPrepares production launches. Use when preparing to deploy to production, or when asking what needs to be in place before shipping. Use when you need a pre-launch checklist, when setting up monitoring, when planning a staged rollout, or when you need a rollback strategy. | Stars: 119k | Updated | Claude CodeCodex | Verified | Install |
| childhoodthedotmack/claude-memBabysit a shipped release after merge, publication, or deployment. Verify published packages, production behavior, installed plugins, and late CI or user reports until the observation window is clean; use babysit for PR readiness before release. | Stars: 118k | Updated | Claude CodeCodex | Verified | Install |