agentleFS
4444J994444J99/a-i--skills/skills/data/data-pipeline-architect/SKILL.md

data-pipeline-architect

Skill17 starsChanged 5 months agoApache-2.0Archived repositoryView on GitHub

Designs ETL/ELT data pipelines with proper extraction, transformation, and loading patterns, including orchestration, error handling, and data quality validation.

5 other projects use a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

6 rules. 0 are common, 6 are distinctive, and 0 are very specific to this project.

  • 0

    Common

    Most of the 3 edited versions we read keep these.

  • 6

    Distinctive

    This author's own choices. Worth a look.

  • 0

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    When target has limited compute

    line 54
  2. Distinctive:

    When transformation reduces data volume significantly

    line 55
  3. Distinctive:

    When sensitive data must be masked before landing

    line 56
  4. Distinctive:

    When raw data preservation is needed

    line 61
  5. Distinctive:

    When transformations change frequently

    line 62
  6. Distinctive:

    Type 2: Add row with validity dates

    line 141

“Common” is measured: kept by at least 6 in 10 of the 3 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Data Pipeline Architect

This skill provides guidance for designing robust, scalable data pipelines that move data reliably from sources to destinations.

Core Competencies

  • ETL vs ELT: Traditional Extract-Transform-Load vs modern Extract-Load-Transform patterns
  • Orchestration: Airflow, Dagster, Prefect, dbt for workflow management
  • Data Quality: Validation, monitoring, lineage tracking
  • Scalability: Batch vs streaming, partitioning, parallelization

Pipeline Design Process

1. Requirements Analysis

To begin pipeline design, gather:

  • Source systems and data formats (APIs, databases, files, streams)
  • Target destinations (data warehouse, lake, lakehouse)
  • Freshness requirements (real-time, hourly, daily)
  • Data volume and velocity estimates
  • Quality and compliance requirements

2. Architecture Selection

Batch Pipelines - For periodic bulk processing:

  • Schedule-driven (hourly, daily, weekly)
  • Higher latency tolerance
  • Simpler error recovery (re-run entire batch)
  • Tools: Airflow, dbt, Spark

Streaming Pipelines - For real-time requirements:

  • Event-driven processing
  • Sub-second to minute latency
  • Complex state management
  • Tools: Kafka, Flink, Spark Streaming

Hybrid Approaches - Lambda or Kappa architecture:

  • Batch layer for completeness
  • Speed layer for low latency
  • Serving layer for queries

3. ETL vs ELT Decision

ETL (Transform before Load):

  • When target has limited compute
  • When transformation reduces data volume significantly
  • When sensitive data must be masked before landing
  • Legacy data warehouse patterns

ELT (Transform after Load):

  • Modern cloud warehouses with cheap compute
  • When raw data preservation is needed
  • When transformations change frequently
  • dbt-style transformations in warehouse

4. Pipeline Components

Extraction Layer:

  • Full extraction vs incremental (CDC, timestamp-based)
  • API pagination and rate limiting
  • Connection pooling and retry logic
  • Schema detection and drift handling

Transformation Layer:

  • Data cleansing and standardization
  • Business logic application
  • Aggregation and denormalization
  • Type casting and null handling

Loading Layer:

  • Upsert strategies (merge, delete+insert)
  • Partitioning schemes (time, hash, range)
  • Index management
  • Transaction boundaries

5. Error Handling Patterns

┌─────────────────────────────────────────────────────────┐
│                    Pipeline Execution                    │
├─────────────────────────────────────────────────────────┤
│  ┌─────────┐    ┌───────────┐    ┌──────────┐          │
│  │ Extract │───▶│ Transform │───▶│   Load   │          │
│  └────┬────┘    └─────┬─────┘    └────┬─────┘          │
│       │               │               │                 │
│       ▼               ▼               ▼                 │
│  ┌─────────┐    ┌───────────┐    ┌──────────┐          │
│  │  Retry  │    │ Dead Letter│    │ Rollback │          │
│  │ w/Backoff│   │   Queue   │    │ Checkpoint│          │
│  └─────────┘    └───────────┘    └──────────┘          │
└─────────────────────────────────────────────────────────┘
  • Retry with backoff: Transient failures (network, rate limits)
  • Dead letter queues: Poison messages that can't be processed
  • Checkpointing: Resume from last successful point
  • Idempotency: Safe to re-run without duplicates

6. Data Quality Framework

Implement checks at each stage:

StageCheck TypeExample
ExtractCompletenessRow count matches source
ExtractFreshnessData timestamp within SLA
TransformValidityValues in expected ranges
TransformUniquenessPrimary keys unique
LoadReconciliationTarget matches source totals
LoadIntegrityForeign keys valid

7. Monitoring and Observability

Essential metrics to track:

  • Pipeline duration and trends
  • Row counts at each stage
  • Error rates and types
  • Data freshness (time since last successful run)
  • Resource utilization

Alert on:

  • SLA breaches (data not fresh)
  • Anomalous row counts (±20% from baseline)
  • Schema changes in sources
  • Repeated failures

Common Patterns

Slowly Changing Dimensions (SCD)

  • Type 1: Overwrite (no history)
  • Type 2: Add row with validity dates
  • Type 3: Previous value column
  • Type 4: History table

Incremental Processing

-- Timestamp-based incremental
SELECT * FROM source
WHERE updated_at > {{ last_run_timestamp }}

-- CDC-based (Change Data Capture)
-- Captures inserts, updates, deletes from transaction log

Idempotent Loads

-- Delete + Insert pattern
DELETE FROM target WHERE date_partition = '2024-01-15';
INSERT INTO target SELECT * FROM staging WHERE date_partition = '2024-01-15';

-- Merge/Upsert pattern
MERGE INTO target t
USING staging s ON t.id = s.id
WHEN MATCHED THEN UPDATE SET ...
WHEN NOT MATCHED THEN INSERT ...

References

  • references/orchestration-patterns.md - Airflow, Dagster, Prefect patterns
  • references/data-quality-checks.md - Validation frameworks and rules
  • references/pipeline-templates.md - Common pipeline architectures

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.