# Build Machine Learning Pipelines

Machine learning projects fail when data, metrics, and deployment choices are unclear. This skill gives structured guidance for model planning, preprocessing, evaluation, and MLOps.

## Install

```bash
npx skillstore add joseph obrien/89jobrien-machine-learning
```

## Metadata

- Status: approved
- Slug: 89jobrien-machine-learning
- Version: 1.0.1
- Author version: 1.0.1
- Skillstore revision: r1
- Version status: valid
- Tree hash: 36ca3887a7ef7adeadf1f38ceda8b641678adb11d94e775c7a36769a6aa5264b
- Author: Joseph OBrien
- GitHub username: 89jobrien
- License: MIT
- Repository: https://github.com/89jobrien/steve/tree/main/steve/skills/machine-learning
- Ref: c4037264bbd363c572662d6154a3ab28f5ca4f53
- Supported tools: Claude, Codex, Claude Code
- Audit status: complete
- Agent install advisory: allowed
- Manual install advisory: allowed
- Artifact signature: available
- Audit attestation: unavailable
- Human verification: not\_verified
- Risk factors: external\_commands
- Quality score: 77
- Quality tier: bronze
- Public page: https://skillstore.pages.dev/skills/89jobrien-machine-learning
- Manifest: https://skillstore.pages.dev/api/skills/89jobrien-machine-learning/manifest

## Capabilities

- Guides ML work from problem definition through production monitoring.
- Maps problem types to suitable metrics and model families.
- Explains preprocessing patterns for missing values, scaling, encoding, outliers, and leakage prevention.
- Covers model training, hyperparameter tuning, cross-validation, and imbalanced data handling.
- Describes serving patterns, versioning, CI/CD checks, monitoring, and retraining triggers.

## Use Cases

- Plan a New ML Project: Define the problem type, success metrics, data split strategy, and first model candidates before implementation begins.
- Prepare Training Data: Choose imputation, encoding, scaling, outlier handling, and leakage prevention steps for a reliable dataset.
- Review Production Readiness: Check serving approach, model versioning, CI/CD tests, monitoring signals, and retraining triggers before launch.

## Prompt Templates

### Choose the ML Approach

```
Use the machine-learning skill to classify my problem type, recommend metrics, and suggest baseline models. My goal is: [describe goal]. My data is: [describe data].
```

### Design Data Preparation

```
Use the machine-learning skill to design preprocessing for this dataset: [columns, types, missing values, target]. Include leakage risks and train-test split guidance.
```

### Build an Evaluation Plan

```
Use the machine-learning skill to create an evaluation plan for [classification, regression, ranking, clustering]. Include primary metrics, validation strategy, and error analysis steps.
```

### Prepare for Production ML

```
Use the machine-learning skill to review production readiness for this model: [model, latency needs, traffic, monitoring plan]. Identify deployment risks and retraining triggers.
```

## Limitations

- It does not train models or inspect datasets without additional user-provided code and data.
- Some referenced files are not included in the package, including model patterns and evaluation guides.
- Examples focus on Python and scikit-learn style workflows.
- It does not provide cloud-specific deployment manifests or infrastructure automation.

## Best Practices

- Define the prediction target, metric, and data split before selecting models.
- Fit preprocessing only on training data, then transform validation and test data.
- Track code version, data version, hyperparameters, metrics, and model artifacts for every experiment.

## Anti Patterns

- Do not tune on the test set or report validation results as final performance.
- Do not use future information or target-derived fields as model inputs.
- Do not deploy without monitoring latency, input drift, prediction drift, and error rates.

## Security Audit

- Audited at: 2026-07-05T12:49:54.649\+00:00
- Summary: All six static findings are false positives caused by Markdown formatting or machine learning terminology. No prompt injection, malicious intent, data exfiltration, or unsafe execution behavior was found in the reviewed files.

## Stats

- Views: 224
- Downloads: 40
- Favorites: 0
- Popularity score: 0