Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add sethdford/claude-skills --skill data-pipeline-designgit clone --depth 1 https://github.com/sethdford/claude-skillsWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/sethdford/claude-skills/data-pipeline-design)<a href="https://agentmods.dev/skills/sethdford/claude-skills/data-pipeline-design"><img src="https://agentmods.dev/badge/skills/sethdford/claude-skills/data-pipeline-design/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/sethdford/claude-skills/data-pipeline-design"><img src="https://agentmods.dev/badge/skills/sethdford/claude-skills/data-pipeline-design.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00037 | $0.00659 |
| Opus 5 | $0.00018 | $0.00329 |
| Sonnet 5 | $0.00007 | $0.00132 |
| Haiku 4.5 | $0.00004 | $0.00066 |
Grade A, and why
data-pipeline-design scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 12d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 48 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Data Pipeline Design
Design robust, maintainable data pipelines that reliably move, transform, and validate data at scale.
Context
You are designing data pipelines (batch or streaming). Plan data flow, transformations, quality gates, failure recovery, and monitoring. Read source systems, target requirements, latency expectations, and volume projections.
Domain Context
Based on modern data engineering practices (Spark, Airflow, Kafka, Beam):
- Batch Pipelines: Scheduled jobs (hourly, daily); high throughput, moderate latency
- Streaming Pipelines: Continuous ingestion; low latency, higher operational complexity
- Micro-batching: Spark Streaming; lower latency than batch, simpler than true streaming
- Orchestration: DAG-based scheduling (Airflow, dbt) for complex multi-stage pipelines
- Observability: Monitor latency, throughput, data quality, freshness
Instructions
-
Choose Processing Model: Batch (daily jobs?) or streaming (realtime features?)? Hybrid (Lambda: batch + streaming for both speed and accuracy)? Consider latency SLA and cost.
-
Design Data Stages: Raw ingestion (as-is from source) → Bronze. Cleansing and normalization → Silver. Business logic and enrichment → Gold. This layered medallion architecture separates concerns.
-
Implement Quality Gates: Validation at each stage. Fail pipeline if data quality drops. Track anomalies: unexpected null rates, value distributions, cardinality changes.
-
Handle Failures and Recovery: Idempotent transformations allow safe retries. Checkpoint state for streaming pipelines; resume from last checkpoint on failure. Use dead-letter queues for unparseable records.
-
Plan Monitoring and Alerting: Track freshness (when was last successful run?), latency (time from source to sink), volume (record counts by stage), error rates. Alert on anomalies and SLA misses.
Anti-Patterns
- No Data Quality Checks: Assume data from source is clean. Result: garbage in, garbage out. Guard: Validate at ingestion; alert on schema changes or anomalies.
- Tightly Coupled Transformations: Pipeline is monolithic script. Result: hard to test, reuse, debug. Guard: Break into modular stages; each stage is independently testable.
- No Checkpoint/Recovery: Assume pipelines always succeed. Result: gaps in data, lost work. Guard: Checkpoint state; design for idempotent retries.
- Ignoring Operational Overhead: Streaming pipelines look simple at 1MB/s, collapse at 1GB/s. Result: unexpected scaling headaches. Guard: Load-test pipelines; plan infrastructure for 10x growth.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 12d ago First seen · 48 lines · 37 tokens per session scan A bf932f7e220c
data-pipeline-design is a skill published in the GitHub repository sethdford/claude-skills (41 stars, last pushed 6mo ago), licensed MIT. It adds 37 tokens to every session and 659 once invoked, about $0.0002 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other skills, from other repositories
cmd_gan_build
GAN pipeline: run gan-planner → gan-generator → gan-evaluator in sequence. Loops until evaluator score ≥ 7.0.
cost-aware-llm-pipeline
Cost optimization patterns for LLM API usage — model routing by task complexity, budget tracking, retry logic, and prompt caching.
regex-vs-llm-structured-text
Decision framework for choosing between regex and LLM when parsing structured text — start with regex, add LLM only for low-confidence edge cases.
foundation-models-on-device
Apple FoundationModels framework for on-device LLM — text generation, guided generation with @Generable, tool calling, and snapshot streaming in iOS 26+.
pytorch-patterns
PyTorch deep learning patterns and best practices for building robust, efficient, and reproducible training pipelines, model architectures, and data loading.
mle-workflow
Production machine-learning engineering workflow for data contracts, reproducible training, model evaluation, deployment, monitoring, and rollback. Use when building, reviewing, or hardening ML systems beyond one-off notebooks.