What does this skill do, and when should you use it?
This skill is for analysts who want to use BigQuery’s built-in machine learning and generative AI capabilities. It helps users write SQL queries that call AI.FORECAST, AI.KEY_DRIVERS, AI.DETECT_ANOMALIES, and AI.GENERATE. Forecasting and anomaly detection use the pre-trained TimesFM model, while content generation uses Gemini models. It is a good fit for teams already working in BigQuery and seeking advanced analysis through SQL.
It guides users in writing BigQuery SQL queries for time-series forecasting, time-series anomaly detection, identifying dimensional segments that drive metric changes, and generating text or other content with Gemini. Its reference directory covers AI.FORECAST, AI.DETECT_ANOMALIES, AI.KEY_DRIVERS, and AI.GENERATE. The source does not say that the skill executes queries or returns actual analytical results itself.
- A data analyst needs forecasts for BigQuery time-series data using the pre-trained TimesFM model.
- An operations or business analyst wants to identify deviations in time-series data.
- A product or growth team needs to find the dimensional segments most responsible for a metric change.
- An analyst wants to generate text or other content directly within BigQuery SQL using Gemini.
- A BigQuery user needs to incorporate machine learning and generative AI operations into an SQL-based analytics workflow.
How do you install this skill?
- Before using AI.GENERATE or the signed-URL image example, verify connection permissions, sensitive-data exposure, model-service data policies, and expected cost.
- Check current BigQuery/Vertex AI regional availability, model versions, and function syntax; the documents are inconsistent about TimesFM versions, and no SQL was executed in this review.
- For mainland-China users, separately verify regional and network reachability of the Google Cloud services; the supplied material provides no Chinese-language or localized-runtime guarantee.
In a client that supports Agent Skills, run npx skills add google/skills, then select the bigquery-ai-ml skill from the repository. The source does not specify the installation directory or a separate per-skill command.
tmp="$(mktemp -d)"
git clone --depth 1 https://github.com/google/skills.git "$tmp"
mkdir -p ~/.claude/skills
cp -R "$tmp/skills/cloud/bigquery-ai-ml" ~/.claude/skills/
rm -rf "$tmp"Generated from the source repository and skill path; it copies only this skill's folder. If the author's install steps above differ, follow those first. To scope it to one project, replace ~/.claude/skills with that project's .claude/skills.
How do you use this skill?
Once installed, send your agent any of these to trigger it:
- Write a BigQuery SQL query using AI.FORECAST for a sales time series and explain the required input fields.
After installation, try a request such as: “Write a BigQuery SQL query using AI.FORECAST for a sales time series and explain the required input fields.” You can also request SQL using AI.DETECT_ANOMALIES, AI.KEY_DRIVERS, or AI.GENERATE. The source does not define fixed trigger phrases, parameter formats, or an execution environment.
What are this skill's strengths and limitations?
- Covers forecasting, anomaly detection, key-driver analysis, and generative AI use cases.
- Uses the pre-trained TimesFM model for forecasting and anomaly detection without custom training.
- Brings Vertex AI capabilities into a BigQuery SQL workflow.
- The repository is licensed under Apache-2.0.
- The skill file provides an overview and reference index but no complete query examples.
- The source does not document permissions, costs, input limitations, or failure handling.
- The source provides no test suite, platform support matrix, or performance metrics for this skill.
- It focuses on BigQuery AI/ML capabilities; core concepts, resource management, CLI usage, and client libraries are covered by the related BigQuery Basics skill.
How does this skill compare with similar options?
Side by side with related skills; every score comes from the same FSRS standard.
| Skill | FS score | Stars | Last updated | License |
|---|---|---|---|---|
| BigQuery AI & ML Analysis Assistant this page ✓ Google · Official | 46 · Use with care | ★ 21k | 12d ago | Apache-2.0 |
| Azure Data Explorer Kusto Analytics ✓ Microsoft · Official | 45 · Not recommended | ★ 2.7k | 3mo ago | MIT |
| CSV/Excel Deep Data Analysis Skill | 47 · Use with care | ★ 20k | 16d ago | MIT |
| BigQuery Asset Impact Analysis ✓ Google · Official | 50 · Use with care | ★ 21k | 12d ago | Apache-2.0 |
| Agent Platform RAG Engine Manager ✓ Google · Official | 49 · Use with care | ★ 21k | 12d ago | Apache-2.0 |
Compared with the related BigQuery Basics skill, this skill focuses on AI/ML and generative AI SQL functions, while BigQuery Basics covers core BigQuery concepts, resource management, the CLI, and client libraries. They are complementary rather than direct substitutes.
How did FollowSkills review this skill?
The documentation exposes data flows involving BigQuery, Vertex AI, tables, prompts, connections, and signed URLs, and includes API status fields. It lacks least-privilege guidance, pre-execution confirmation, sensitive-data handling, cost controls, external-model governance, and rollback guidance, so points are deducted. No credential theft, covert exfiltration, or destructive default is evident.
The four reference files provide broadly consistent syntax, parameters, output fields, examples, and some failure-status guidance, including a three-data-point minimum for some functions. There is no test suite, CI evidence, reproducibility proof, or systematic abnormal-input handling, and TimesFM 2.0 and 2.5 are inconsistent across defaults and examples, so points are deducted. Static review caps this at 10.
The scenarios are clear: forecasting, anomaly detection, key-driver analysis, and generative AI, with input/output examples. However, trigger conditions, non-fit boundaries, permission and regional prerequisites, Chinese support, and mainland-China reachability are not documented. BigQuery and Vertex AI availability may depend on account, region, and network conditions, so points are deducted.
The main document offers progressive disclosure through four linked references, while the repository README supplies installation, support, contribution, Apache-2.0 licensing, and active-development information. The skill lacks dedicated dependency notes, compatibility/versioning, a changelog, FAQs, explicit limitations, and clearly assigned maintenance ownership, so points are deducted.
Each topic includes reusable SQL syntax, parameter tables, output schemas, and examples that address the core tasks. The examples do not adequately cover connection setup, data quality, cost, model-version, or production-safety boundaries, and none is executed or independently verified here, so user debugging and review remain necessary. Static review caps this at 7.
The references provide concrete function signatures, fields, and status columns suitable for text-level audit. They contain no official citations, committed tests, CI results, or third-party execution evidence, so key claims cannot be independently corroborated from the supplied static material. Static review caps this at 5.
Open a dimension to read why it scored that way
Evidence confidence：Low — Mostly static review, author material or a limited demo; useful for discovery, not high-risk decisions.
See the full review method →