timesfm-forecasting
Zero-shot time series forecasting with Google's TimesFM foundation model. Use this skill when forecasting ANY univariate time series — sales, sensor readings, stock prices, energy demand, patient vitals, weather, or scientific measurements — without training a custom model. Automatically checks system RAM/GPU before loading the model, supports CSV/DataFrame/array inputs, and returns point forecasts with calibrated prediction intervals. Includes a preflight system checker script that MUST be run before first use to verify the machine can load the model. For classical statistical time series m
npx skills add foryourhealth111-pixel/Vibe-Skills --skill timesfm-forecasting --agent claude-code
Same command for any agent — swap --agent for codex, cursor, copilot.
Weekly change comes from our own snapshots, not the repository page — it measures attention, not adoption.
What it does
Forecasting of any univariate time series using a TimesFM foundation model in a zero-shot setup. Provides point forecasts plus calibrated quantile prediction intervals, and supports inputs as CSV, DataFrame, or arrays. Includes a mandatory preflight system checker that must be run before first use to verify RAM, GPU, disk space, Python version, and existing installations. If checks pass, the agent loads the TimesFM model and runs forecasts with configurable options.
How it works
- The skill requires running a preflight script:
python scripts/check_system.pyto verify RAM, GPU availability, disk space, Python version, and existing installations fortimesfmandtorch. - After the preflight succeeds, install TimesFM (e.g.,
uv pip install timesfm[torch]orpip install timesfm[torch]) and the appropriate PyTorch backend per hardware as described. - The agent loads a TimesFM model via
TimesFM_2p5_200M_torch.from_pretrained("google/timesfm-2.5-200m-pytorch")and compiles it with atimesfm.ForecastConfigobject, specifying parameters such asmax_context,max_horizon,normalize_inputs,use_continuous_quantile_head,force_flip_invariance,infer_is_positive, andfix_quantile_crossing. - Forecasts are produced with
model.forecast(horizon=..., inputs=...), returningpointandquantileswherequantilesencodes 10 quantile slices and the 5th slice corresponds to the median. - Optional capabilities include
forecast_with_covariates()for covariates, and usage of quantile intervals for anomaly detection by comparing actuals to lower/upper bounds.
When to use it
- When forecasting any univariate time series with zero training required.
- When probabilistic forecasts with calibrated intervals are desired.
- When input data can be provided as 1-D arrays, CSV, or DataFrames and batch forecasting of many series is needed.
- When you prefer a foundation-model approach over classical ARIMA/ETS parameters.
What it can touch
- The skill uses the TimesFM model and PyTorch backend; it references
timesfmandtorchinstallations and model weights downloaded on-demand from HuggingFace and cached in~/.cache/huggingface/. - It supports inputs via CSV reading and numpy arrays; covariates can be provided via
forecast_with_covariates()with appropriatexregconfigurations.
Caveats
- The preflight script MUST be run before first use to verify machine readiness.
- TimesFM model weights (~800 MB) are downloaded on-demand and not stored in the repository.
- This skill is Apache-2.0 licensed.
# TimesFM Forecasting ## Routing Boundary Use this skill only for TimesFM, zero-shot forecasting, foundation forecasting, forecast horizons, prediction intervals, or TimesFM-specific time-series pipelines. Do not use it for generic business forecasting, ARIMA baselines, tabular regression, ordinary scikit-learn modeling, or exploratory time-series analysis without TimesFM/foundation-model signals.
What does the timesfm-forecasting skill do?
Zero-shot time series forecasting with Google's TimesFM foundation model. Use this skill when forecasting ANY univariate time series — sales, sensor readings, stock prices, energy demand, patient vitals, weather, or scientific measurements — without training a custom model. Automatically checks system RAM/GPU before loading the model, supports CSV/DataFrame/array inputs, and returns point forecasts with calibrated prediction intervals. Includes a preflight system checker script that MUST be run before first use to verify the machine can load the model. For classical statistical time series m
How do I install it?
Run `npx skills add foryourhealth111-pixel/Vibe-Skills --skill timesfm-forecasting --agent claude-code` — it drops the skill into your project so the agent can pick it up. Swap the --agent value for codex, cursor or copilot if you use one of those.
Where does this skill come from?
From foryourhealth111-pixel/Vibe-Skills, a repository with 2,593 stars. We read it straight from the repository tree rather than a submitted listing, so what you see here is what is actually published.
Is a popular skill a good skill?
Not necessarily. Stars measure attention, not adoption — a repository can trend for a week and be abandoned. That is why we show the weekly change from our own snapshots next to the total, instead of a single flattering number.