Feature Store And Ml Data Pipelines
Guides agents through machine-learning data pipelines and feature serving workflows. Use when designing feature generation, offline and online consistency, training-serving parity, point-in-time correctness, or ML-oriented data product contracts.
Install
$ agentstack add skill-vaquarkhan-data-engineering-agent-skills-feature-store-and-ml-data-pipelines ✓ scanned · ✓ verified — works with Claude Code, Cursor, and more.
Security review
✓ PassedNo issues found. Passed automated security review. · v0.1.0 How review works →
- ✓ Prompt-injection patterns
- ✓ Secret / credential exfiltration
- ✓ Dangerous shell & filesystem operations
- ✓ Untrusted network calls
- ✓ Known-malicious package signatures
What it can access
- ✓ Network access No
- ✓ Filesystem access No
- ✓ Shell / process execution No
- ✓ Environment & secrets No
- ✓ Dynamic code execution No
From automated source analysis of v0.1.0. “Used” means the capability is present in the source — more access means more to trust, not that it’s unsafe.
About
Feature Store And ML Data Pipelines
Overview
Use this skill when the platform must support model training and inference safely. It helps agents design feature generation, point-in-time correctness, serving parity, and operational contracts for ML-focused data products.
When to Use
- building training datasets
- designing feature stores or reusable features
- supporting online and offline feature access
- preventing leakage and training-serving mismatch
- publishing model-ready data products
Do not treat feature pipelines as ordinary marts with different names. ML pipelines have different correctness risks.
Workflow
- Define the feature contract.
Include:
- entity key
- feature meaning
- update cadence
- online or offline use
- freshness expectation
- Protect point-in-time correctness.
Training data must only include information available at prediction time.
- Align offline and online logic.
Reuse definitions and validation wherever possible to prevent training-serving drift.
- Define feature lifecycle and ownership.
Clarify:
- producer
- consumers
- deprecation path
- quality monitoring
- Validate operational behavior.
Models break when stale or missing features silently propagate.
Common Rationalizations
| Rationalization | Reality | | --- | --- | | "We can use the latest value for training." | That often introduces leakage and overstates model performance. | | "Online parity is a model-team problem." | Feature consistency is a data pipeline responsibility too. | | "Features are internal, so contracts are unnecessary." | Unclear feature meaning leads to misuse and drift. |
Red Flags
- no point-in-time logic is defined
- offline and online definitions diverge
- stale features are not monitored
- feature ownership is unclear
Verification
- [ ] Feature meaning, keys, and freshness are documented
- [ ] Point-in-time correctness is protected
- [ ] Offline and online parity expectations are explicit
- [ ] Monitoring exists for stale, missing, or drifting features
Source & license
This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.
- Author: vaquarkhan
- Source: vaquarkhan/data-engineering-agent-skills
- License: MIT
- Homepage: https://vaquarkhan.github.io/data-engineering-agent-skills/
Install and usage instructions live in the source repository linked above.
Reviews
No reviews yet — be the first.
Write a review
Versions
- v0.1.0 Imported from the upstream source.