CtrlK
BlogDocsLog inGet started
Tessl Logo
SkillPublic

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. Supports both basic forecasting and advanced covariate forecasting (XReg) with dynamic and static exogenous variables. Automatically checks system RAM/GPU before loading the model, validates dataset fit before processing, 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 and handle your specific dataset.

Quality

90%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide
SecuritybySnyk

Passed

No findings from the security scan

SKILL.md
Quality
Evals
Security

Quality

Content

81%Weight 40%Scale 1-5

Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.

A highly actionable, workflow-complete skill body: executable code for every common case, a mandatory preflight gate, post-task checklist, and runnable validation commands. Its weaknesses are verbosity — duplicated check_system.py and model-version sections inflate the file — and references to examples/ directories that do not exist in the bundle.

Suggestions

Consolidate the check_system.py documentation: the 'Mandatory Preflight' and 'Available Scripts' sections repeat the same commands and five-item check list verbatim — keep one canonical section and cross-reference it.

Merge the two model-version tables ('Hardware Requirements by Model Version' and 'Model Versions') into a single table, and drop the inline ForecastConfig parameter table in favor of the existing references/api_reference.md to cut duplicated content.

Either ship the three examples/ directories referenced in the Examples section and validation commands, or remove/trim those references so navigation does not point at files absent from the bundle.

DimensionReasoningScore

Conciseness

Mostly dense, skill-specific material (quantile index tables, memory formula, ForecastConfig tuning table), but there is real duplication: the check_system.py invocation and its "What it checks" list appear nearly verbatim in both "Mandatory Preflight" and "Available Scripts"; model-version info appears in two separate tables; and the decorative mermaid flowchart plus "Example Outputs" blocks pad the file. Not 4: the duplicated sections and ~500-line length mean more than minor trimming is possible; not 2: the bulk is TimesFM-specific knowledge Claude does not already have, not generic explanation.

3 / 5

Actionability

Fully executable, copy-paste-ready code throughout: the minimal example, covariate call, anomaly detection snippet, batch forecasting, evaluation metrics, plotting, and CLI commands for both scripts. The Quality Checklist and Common Mistakes give concrete, checkable guidance (e.g., "quant_fc[..., 0] is the mean, not q0... Define: IDX_Q10, IDX_Q90 = 1, 9"). Not 4: the examples cover the common cases completely, with specific commands, expected shapes, and acceptance criteria.

5 / 5

Workflow Clarity

Clear sequence (system check → install → forecast → checklist → validation) with explicit validation checkpoints: "CRITICAL — ALWAYS run the system checker before loading the model", the post-task Quality Checklist, runnable regression assertions in "Validation & Verification", and error-recovery guidance ("Try: context_length=512 or process in chunks of 50 series"). Not 4: batch operations are covered by both the checklist and executable validation commands, satisfying the feedback-loop expectation for batch workflows.

5 / 5

Progressive Disclosure

Good structure: the three real references/ files are one level deep and clearly signaled in a table with contents, scripts are documented, and only an API summary is inline with the bulk in references/api_reference.md. Not 5: the body cites examples/global-temperature/, examples/anomaly-detection/, and examples/covariates-forecasting/ (including run commands and validation paths) that are not present in the bundle, and some content (ForecastConfig parameter table, hardware tables) is duplicated inline despite existing in the reference files.

4 / 5

Total

17

/

20

Passed

Description

96%Weight 40%Scale 1-5

Based on the skill's description, can an agent find and select it at the right time? Clear, specific descriptions lead to better discovery.

A strong description: concrete capability list, explicit "Use this skill when" trigger clause with natural domain keywords, consistent third-person voice, and a clear foundation-model niche. Minor deductions for slight redundancy (the system checker is described twice in different words) and broad ANY-forecasting claims that create minor overlap risk with adjacent forecasting skills.

DimensionReasoningScore

Specificity

Lists multiple concrete actions — "covariate forecasting (XReg) with dynamic and static exogenous variables", "checks system RAM/GPU", "supports CSV/DataFrame/array inputs", "returns point forecasts with calibrated prediction intervals" — with comprehensive coverage of the skill's capabilities. Not 4: coverage spans basic forecasting, covariates, system checking, input formats, and output format, matching the comprehensive anchor; the third-person voice is consistent throughout.

5 / 5

Completeness

Explicitly answers both: what ("Zero-shot time series forecasting with Google's TimesFM foundation model... returns point forecasts with calibrated prediction intervals") and when ("Use this skill when forecasting ANY univariate time series — sales, sensor readings, stock prices..."). Not 4: the "when" clause is explicit and enumerates concrete trigger scenarios rather than being merely present.

5 / 5

Trigger Term Quality

Natural user phrases are well covered: "forecasting", "sales", "sensor readings", "stock prices", "energy demand", "patient vitals", "weather", "scientific measurements", plus "CSV". Not 4: this goes beyond a few keywords to include domain synonyms across the whole range of series a user would name when asking for a forecast.

5 / 5

Distinctiveness Conflict Risk

The TimesFM/zero-shot/foundation-model framing carves a clear niche, but "forecasting ANY univariate time series... without training a custom model" claims broad territory that overlaps with generic forecasting or ML-skills alternatives. Not 5: a closely related classical-forecasting skill would trigger on many of the same natural phrases; not 3: the TimesFM-specific anchoring keeps overlap limited to near-neighbor skills.

4 / 5

Total

19

/

20

Passed

Validation

93%

Checks the skill against the spec for correct structure and formatting. All validation checks must pass before discovery and implementation can be scored.

Validation — 15 / 16 Passed

Validation for skill structure

CriteriaDescriptionResult

skill_md_line_count

SKILL.md is long (512 lines); consider splitting into references/ and linking

Warning

Total

15

/

16

Passed

Repository

google-research/timesfm

Path

timesfm-forecasting/SKILL.md

Reviewed

Table of Contents

Is this your skill?

If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.