# Can google/skills Be Used for Machine Learning Tasks? A Complete Guide to BigQuery AI & ML Skills

> Discover how to leverage google/skills for machine learning tasks with BigQuery AI & ML. Perform forecasting, classification, semantic search, text generation, and more using SQL.

- Repository: [Google/skills](https://github.com/google/skills)
- Tags: deep-dive
- Published: 2026-08-14

---

**Yes, google/skills can be used for machine learning tasks through the BigQuery AI & ML skill, which wraps BigQuery ML and Vertex AI functions for forecasting, classification, anomaly detection, semantic search, vector search, text generation, and embedding generation—all executable via SQL.**

The google/skills repository provides declarative "Skill" definitions that describe how to interact with Google Cloud services. As implemented in google/skills, the **BigQuery AI & ML** skill (category `AiAndMachineLearning`) enables you to run sophisticated ML workloads without provisioning separate compute resources. This skill translates declarative metadata into native BigQuery AI function calls.

## How the BigQuery AI & ML Skill Works

The skill architecture follows a three-layer pattern:

- **Skill Definition (YAML front-matter)** – declares the skill name, category, and description in [`SKILL.md`](https://github.com/google/skills/blob/main/SKILL.md)
- **Reference Files** – individual markdown files documenting each AI function's SQL syntax, parameters, and examples
- **Runtime Execution** – the Skills framework expands metadata into concrete BigQuery requests executed by BigQuery's managed service

Because the skill is **declarative**, you can integrate ML capabilities into Automation scripts, Cloud Build steps, and AI Agents without writing custom client-library code.

## Machine Learning Capabilities in google/skills

The BigQuery AI & ML skill supports both **classic statistical ML** and **modern generative AI** workloads. Below are practical implementations with source-verified SQL syntax.

### Time-Series Forecasting with AI.FORECAST

Use `AI.FORECAST` for demand planning, capacity management, and trend prediction.

```sql
-- Forecast the next 30 days of sales using a simple linear model
SELECT
  *
FROM
  ML.FORECAST_MODEL(
    MODEL `myproject.dataset.sales_forecast`,
    STRUCT(30 AS horizon, 'daily' AS time_granularity)
  );

```

*Source:* [[`ai_forecast.md`](https://github.com/google/skills/blob/main/ai_forecast.md)](https://github.com/google/skills/blob/main/skills/cloud/bigquery-ai-ml/references/ai_forecast.md)

### Text Generation with AI.GENERATE

Generate content, summaries, or structured outputs using hosted LLMs directly in SQL.

```sql
SELECT
  AI.GENERATE(
    prompt => 'Write a short blog post about the benefits of serverless computing.',
    temperature => 0.7,
    max_output_tokens => 256
  ) AS generated_text;

```

*Source:* [[`ai_generate.md`](https://github.com/google/skills/blob/main/ai_generate.md)](https://github.com/google/skills/blob/main/skills/cloud/bigquery-ai-ml/references/ai_generate.md)

### Semantic Search with AI.SEARCH

Implement **retrieval-augmented generation (RAG)** pipelines by matching queries to embedded documents.

```sql
SELECT
  *
FROM
  `myproject.dataset.documents`
WHERE
  AI.SEARCH(
    text => 'machine learning pipelines',
    vector_column => embedding_vector,
    top_k => 5
  );

```

*Source:* [[`ai_search.md`](https://github.com/google/skills/blob/main/ai_search.md)](https://github.com/google/skills/blob/main/skills/cloud/bigquery-ai-ml/references/ai_search.md)

### Anomaly Detection with AI.DETECT_ANOMALIES

Monitor metrics and flag outliers in time-series data without manual threshold tuning.

```sql
SELECT
  *
FROM
  AI.DETECT_ANOMALIES(
    TABLE `myproject.dataset.metric_timeseries`,
    time_column => 'timestamp',
    value_column => 'value',
    sensitivity => 0.8
  );

```

*Source:* [[`ai_detect_anomalies.md`](https://github.com/google/skills/blob/main/ai_detect_anomalies.md)](https://github.com/google/skills/blob/main/skills/cloud/bigquery-ai-ml/references/ai_detect_anomalies.md)

### Vector Search with VECTOR_SEARCH

Execute **hybrid retrieval** combining vector similarity with structured filters.

```sql
SELECT
  *
FROM
  VECTOR_SEARCH(
    TABLE `myproject.dataset.product_embeddings`,
    query_vector => AI.GENERATE_EMBEDDING('smartphone with great camera'),
    top_k => 10
  );

```

*Source:* [[`vector_search.md`](https://github.com/google/skills/blob/main/vector_search.md)](https://github.com/google/skills/blob/main/skills/cloud/bigquery-ai-ml/references/vector_search.md)

## Key Source Files for ML Development

According to the google/skills source code, these files define and document machine learning capabilities:

| Path | Purpose |
|------|---------|
| [[`skills/cloud/bigquery-ai-ml/SKILL.md`](https://github.com/google/skills/blob/main/skills/cloud/bigquery-ai-ml/SKILL.md)](https://github.com/google/skills/blob/main/skills/cloud/bigquery-ai-ml/SKILL.md) | Core skill definition declaring `AiAndMachineLearning` category |
| [`skills/cloud/bigquery-ai-ml/references/ai_forecast.md`](https://github.com/google/skills/blob/main/skills/cloud/bigquery-ai-ml/references/ai_forecast.md) | Time-series forecasting syntax and examples |
| [`skills/cloud/bigquery-ai-ml/references/ai_generate.md`](https://github.com/google/skills/blob/main/skills/cloud/bigquery-ai-ml/references/ai_generate.md) | LLM text generation parameters |
| [`skills/cloud/bigquery-ai-ml/references/ai_search.md`](https://github.com/google/skills/blob/main/skills/cloud/bigquery-ai-ml/references/ai_search.md) | Semantic search implementation |
| [`skills/cloud/bigquery-ai-ml/references/ai_detect_anomalies.md`](https://github.com/google/skills/blob/main/skills/cloud/bigquery-ai-ml/references/ai_detect_anomalies.md) | Anomaly detection configuration |
| [`skills/cloud/bigquery-ai-ml/references/vector_search.md`](https://github.com/google/skills/blob/main/skills/cloud/bigquery-ai-ml/references/vector_search.md) | Vector search and embedding retrieval |
| [`skills/cloud/bigquery-bigframes/SKILL.md`](https://github.com/google/skills/blob/main/skills/cloud/bigquery-bigframes/SKILL.md) | Higher-level ML utilities via BigQuery BigFrames |
| [`skills/cloud/bigquery-bigframes/references/linear_regression.md`](https://github.com/google/skills/blob/main/skills/cloud/bigquery-bigframes/references/linear_regression.md) | Classic linear regression with BigFrames |
| [`skills/cloud/bigquery-bigframes/references/logistic_regression.md`](https://github.com/google/skills/blob/main/skills/cloud/bigquery-bigframes/references/logistic_regression.md) | Classification with BigFrames |

## Integration Patterns for ML Workflows

The `AiAndMachineLearning` category enables tooling to **automatically discover ML-capable skills**. This supports several integration patterns:

- **Skills CLI** – invoke ML functions directly from command-line workflows
- **Cloud Build** – embed model training or inference into CI/CD pipelines
- **AI Agents** – let autonomous systems discover and execute ML capabilities
- **Automation Platforms** – trigger forecasts or anomaly detection on schedules

All compute executes within BigQuery's managed service—no separate Vertex AI or Compute Engine provisioning required.

## Summary

- **google/skills supports machine learning tasks** through the BigQuery AI & ML skill category
- **Declarative YAML definitions** in [`SKILL.md`](https://github.com/google/skills/blob/main/SKILL.md) enable discovery and orchestration without custom code
- **Native BigQuery SQL functions** cover forecasting (`AI.FORECAST`), generation (`AI.GENERATE`), search (`AI.SEARCH`, `VECTOR_SEARCH`), and anomaly detection (`AI.DETECT_ANOMALIES`)
- **BigFrames skills** provide additional classic ML utilities for regression and classification
- **Zero compute provisioning**—workloads run entirely within BigQuery's managed infrastructure

## Frequently Asked Questions

### What types of machine learning does google/skills support?

google/skills supports **time-series forecasting**, **classification**, **anomaly detection**, **semantic search**, **vector search**, **text generation**, and **embedding generation**. These capabilities map directly to BigQuery ML and Vertex AI functions exposed through SQL syntax in the BigQuery AI & ML skill.

### Do I need to provision Vertex AI infrastructure to use ML skills?

No. The BigQuery AI & ML skill executes entirely within **BigQuery's managed service**. When you invoke a skill, the framework expands metadata into a BigQuery request. Heavy computation—including model inference and vector operations—runs on BigQuery's infrastructure without requiring separate Vertex AI or Compute Engine resources.

### How do I discover ML-capable skills programmatically?

Skills are tagged with categories in their YAML front-matter. The BigQuery AI & ML skill uses category `AiAndMachineLearning`. Tools can scan [`SKILL.md`](https://github.com/google/skills/blob/main/SKILL.md) files and filter by this category to automatically enumerate all available machine learning capabilities in a google/skills deployment.

### Can I use google/skills for traditional statistical ML, not just generative AI?

Yes. Beyond generative AI functions, the **BigQuery BigFrames skill** ([`skills/cloud/bigquery-bigframes/SKILL.md`](https://github.com/google/skills/blob/main/skills/cloud/bigquery-bigframes/SKILL.md)) provides classic ML utilities including linear regression and logistic regression. This skill offers higher-level abstractions for statistical modeling while maintaining the same declarative, SQL-based execution model.