Skip to content

mikeparcewski/wicked-garden / skills/data

wicked-garden-data

Verified

Data engineering, analysis, and ML toolkit. One skill, routed sub-actions: analyze (interactive EDA on a CSV/Excel/data file), profile / validate / quality (schema-level data engineering ops), ml review / ml pipeline (model review + training-pipeline design), pipeline design / pipeline review (ETL architecture), ontology (recommend Schema.org, Dublin Core, DCAT, FOAF, GoodRelations, SKOS, or a custom shape for a dataset), shapes (a live source's data shapes, proposed for review). Use when: "analyze this CSV/Excel/data file" / "explore this dataset" / one-off data exploration; "profile this dataset" / "validate data against a schema" / "generate a data quality report"; "review this ML model" / "design an ML training pipeline"; "design a data pipeline" / "review this ETL"; "recommend an ontology for this data" / "map columns to a public ontology"; "propose data shapes for <source>"; or any former /wicked-garden:data:* invocation. Delegated deep passes: the `wicked-garden-data-engineer` worker skill.

FileKindSize
SKILL.mdskill md8.0 KB
refs/analysis-templates.mdmarkdown3.3 KB
refs/analyze.mdmarkdown5.1 KB
refs/data-examples.mdmarkdown4.9 KB
refs/data.mdmarkdown3.8 KB
refs/ml-templates.mdmarkdown4.0 KB
refs/ml.mdmarkdown5.7 KB
refs/pipeline-templates.mdmarkdown4.6 KB
refs/pipeline.mdmarkdown4.8 KB
refs/shapes.mdmarkdown6.1 KB

refs/ml.md

View on GitHub
# ML Engineering Rubric — review / pipeline

Apply this inline. Caller has pre-gathered model files (for review) or
captured task-type (for pipeline design).

## review

```markdown
## ML Engineering Assessment

**Model**: {name}
**Type**: {classification|regression|ranking|other}
**Status**: {Development|Staging|Production}

### Summary
{2-3 sentence overview}

### Architecture Review
- **Problem Type**: {clear definition}
- **Model Type**: {algorithm/architecture}
- **Justification**: {why this approach}
- **Alternatives Considered**: {other options}

### Data & Features
- **Training Size**: {rows}
- **Feature Count**: {count}
- **Quality**: {assessment}

| Check | Status | Notes |
|-------|--------|-------|
| No data leakage (target in features) | {pass/fail} | |
| Label quality | {pass/fail} | |
| Class imbalance handled | {pass/fail} | |
| Missing values handled | {pass/fail} | |
| Categorical encoding defined | {pass/fail} | |
| Feature importance tracked | {pass/fail} | |

### Evaluation Strategy
- **Metrics**: {precision, recall, RMSE, etc.}
- **Validation**: {holdout / k-fold / time-based}
- **Success Criteria**: {minimum acceptable performance}

### Deployment Readiness
- [ ] Inference latency acceptable
- [ ] Model size appropriate for deployment target
- [ ] Feature pipeline reproducible in production
- [ ] Graceful fallback if model fails
- [ ] A/B testing framework ready
- [ ] Monitoring and logging instrumented
- [ ] Rollback plan documented

### Recommendations
| Priority | Recommendation | Impact | Effort |
|----------|----------------|--------|--------|
| P1 | {critical item} | HIGH | {S/M/L} |

**Recommendation**: {APPROVE|CONDITIONAL|NEEDS REWORK}
**Confidence**: {HIGH|MEDIUM|LOW}
```

## pipeline (training pipeline design)

Problem setup → data/feature stages → model arch → evaluation → tuning → deployment → monitoring.

````markdown
## Training Pipeline: {model_name}

### Problem Statement
- **Type**: {Classification|Regression|Ranking|Other}
- **Target**: {what is being predicted}
- **Business Metric**: {how success is measured}
- **Baseline**: {minimum viable baseline established?}

### Data Flow
1. **Source**: {where data comes from}
2. **Sampling**: {train/val/test split strategy — e.g. 70/15/15}
3. **Features**: {transformation pipeline, feature store integration}
4. **Training**: {model training config, hyperparameter search}
5. **Validation**: {holdout / cross-validation approach}
6. **Registry**: {versioning scheme, metadata, promotion criteria}

### Configuration (sample)
```yaml
data:
  source: {path}
  train_split: 0.7  val_split: 0.15  test_split: 0.15
features:
  categorical: [col1, col2]
  numerical: [col3, col4]
  target: label
model:
  type: {algorithm}
  key_params: {}
evaluation:
  metric: {f1_score|rmse|ndcg}
  threshold: {value}
```

### Monitoring Plan
- **Drift detection**: {feature/target drift strategy}
- **Performance tracking**: {metrics logged}
- **Retraining trigger**: {when to retrain}

### MLOps Principles
- **Reproducible**: version data, code, and models.
- **Monitored**: track performance continuously.
- **Automated**: CI/CD for ML pipelines.
- **Tested**: unit tests for features, integration tests for pipeline.
- **Documented**: model cards for transparency.
````

## ML development workflow (7 steps)

### 1. Problem definition
- [ ] Problem type clear (classification/regression/ranking)
- [ ] Success metric defined
- [ ] Baseline established
- [ ] Data availability confirmed
- [ ] Inference requirements understood (latency/throughput)

### 2. Data assessment
- [ ] Sufficient volume (>1000x features)
- [ ] Labels accurate and consistent
- [ ] Features available at inference time
- [ ] No data leakage from target
- [ ] Class balance acceptable

### 3. Feature engineering
**Good features are**: predictive, available at inference, clean (no leakage),
interpretable.

**Common transformations**:
- Numeric: scaling, log transform
- Categorical: one-hot, target encoding
- Time: extract components, cyclical encoding
- Aggregations: rolling windows, user stats

### 4. Model selection
| Data Size | Structured | Recommendation |
|-----------|------------|----------------|
| <10K rows | Yes | Linear/Simple tree |
| 10K-1M | Yes | GradientBoosting (XGBoost/LightGBM) |
| >1M | Yes | Deep learning possible |
| Any | Images/Text | Deep learning |

### 5. Training & evaluation
**Split strategy**: random (if i.i.d.), time-based (if time series),
cross-validation (robust).

**Key metrics**:
- **Classification**: Accuracy, Precision, Recall, F1, AUC
- **Regression**: RMSE, MAE, R²

### 6. Hyperparameter tuning
- **Grid search**: exhaustive, slow.
- **Random search**: more efficient.
- **Bayesian optimization**: most efficient.

### 7. Deployment
**Patterns**: batch scoring, REST API, streaming. Apply the Deployment
Readiness checklist from the review rubric above.

## ML monitoring

- **Model performance**: prediction accuracy, distribution shifts, error rate by segment.
- **Data quality**: feature distributions, missing rates, cardinality changes.
- **System health**: latency (p50, p95, p99), throughput, memory.

## Best practices

- **Always baseline**: start simple, measure improvement, justify complexity.
- **Avoid leakage**: use only past data, split before processing.
- **Monitor production**: track predictions, detect drift, plan retraining.
- **Document everything**: architecture, features, training data, results.

## Common pitfalls

- Training on future data (leakage).
- Overfitting to validation set.
- Ignoring class imbalance.
- Not testing inference performance.
- Deploying without monitoring.

## Detailed templates

See [ml-templates.md](ml-templates.md) for the scored 5-section model review
report, model card, and experiment log templates.
writing-skillsobra/superpowersUse when creating new skills, editing existing skills, or verifying skills work before deploymentStars: 322kUpdated Claude CodeCodexVerifiedInstall
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: 198kUpdated Claude CodeCodexVerifiedInstall
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: 119kUpdated Claude CodeCodexVerifiedInstall
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: 118kUpdated Claude CodeCodexVerifiedInstall