giskard-tests

Test ML models with Giskard's scan() vulnerability detector + test catalog (performance, robustness, fairness, data leakage, ethical issues) for tabular and NLP models. Wrap a prediction function in giskard.Model + a DataFrame in giskard.Dataset; emit test suites that pass/fail in CI. Use when a trained tabular or NLP model is about to ship with no test suite of its own, or when a feature-engineering or hyperparameter change needs a pre-merge scan for newly introduced vulnerabilities.

Install with skills.sh (any agent)

npx skills add testland/qa --skill giskard-tests

View source

giskard-tests

Giskard wraps any prediction function and DataFrame, then runs a scan() that surfaces "performance biases, unrobustness, data leakage, stochasticity, underconfidence, ethical issues" per the Giskard tabular quickstart (opens in new window).

When to use

Pre-deployment model audit (fairness + robustness scan).
Pre-merge gate after a feature-engineering or hyperparameter change - re-run scan, fail PR if new vulnerabilities surface.
Generating an initial test suite for a model with no existing tests.

Step 1 - Install

pip install giskard --upgrade

Per the Giskard tabular quickstart (opens in new window).

Step 2 - Wrap the dataset

from giskard import Dataset

giskard_dataset = Dataset(
    df=raw_data,
    target=TARGET_COLUMN,
    name="Titanic dataset",
    cat_columns=CATEGORICAL_COLUMNS,
)

cat_columns matters - Giskard treats categoricals differently for slicing + drift detection.

Step 3 - Wrap the model

from giskard import Model
import numpy as np
import pandas as pd

def prediction_function(df: pd.DataFrame) -> np.ndarray:
    preprocessed_df = preprocessing_function(df)
    return classifier.predict_proba(preprocessed_df)

giskard_model = Model(
    model=prediction_function,
    model_type="classification",
    name="Titanic model",
    classification_labels=classifier.classes_,
    feature_names=FEATURE_NAMES,
)

The prediction_function returns probabilities (not class labels) for classification - required by Giskard's calibration checks.

Step 4 - Scan for vulnerabilities

from giskard import scan

results = scan(giskard_model, giskard_dataset)
results.to_html("scan_report.html")

Per the Giskard tabular quickstart (opens in new window), scan covers categories: performance bias, unrobustness, data leakage, stochasticity, underconfidence, ethical issues. HTML report is artifact-friendly for CI.

Step 5 - Generate a test suite from scan

test_suite = results.generate_test_suite("My first test suite")
suite_results = test_suite.run()

if not suite_results.passed:
    raise SystemExit("Giskard test suite failed; see report")

Step 6 - Add specific tests from catalog

from giskard import testing

test_suite.add_test(
    testing.test_f1(
        model=giskard_model,
        dataset=giskard_dataset,
        threshold=0.7,
    )
)

# Slicing test: F1 must hold on a subset
female_slice = giskard_dataset.slice(lambda df: df[df.sex == "female"])
test_suite.add_test(
    testing.test_f1(
        model=giskard_model,
        dataset=female_slice,
        threshold=0.65,
    )
)

test_suite.run()

Catalog includes test_f1, test_accuracy, test_recall, test_drift_*, metamorphic transformations. Reference the Giskard tabular quickstart (opens in new window) for the current full list.

Step 7 - CI integration

- name: Giskard scan
  run: |
    python ml/giskard_scan.py
    # Script raises SystemExit on failure
- name: Upload Giskard report
  uses: actions/upload-artifact@v4
  with:
    name: giskard-report
    path: scan_report.html

Anti-patterns

Anti-pattern	Why it fails	Fix
Skip `cat_columns` parameter	Categorical features treated as numeric; bogus drift	Always pass `cat_columns` (Step 2)
Wrap a `predict()` (classes) instead of `predict_proba()` (probs)	Calibration tests cannot run	Use `predict_proba` for classification (Step 3)
Run scan once, don't add to suite	One-off finding never re-checked	Generate suite from scan (Step 5); CI gates re-run
Block CI on every minor scan finding	Noise; team disables Giskard	Set per-test threshold; gate on critical+major only
Reuse training dataset for scan	False sense of robustness; scan needs unseen data	Use held-out test split

Limitations

Giskard's NLP scan covers fewer languages than tabular. Verify language support in the Giskard tabular quickstart (opens in new window) sister pages.
scan() is non-deterministic with stochastic models; pin random_state in the model and Giskard config for reproducible CI.

References

Giskard tabular quickstart (opens in new window) - Model/Dataset wrapping, scan, suite generation, test catalog

Related skills

alibi-explainability

Generates model explanations with Alibi Explain - Anchors, Integrated Gradients, Kernel/Tree SHAP, ALE, Counterfactual Instances. Wires explainer.fit + explainer.explain into model-evaluation pipelines so that every flagged prediction ships with a "why" record auditors can reason about. Use when a model decision must be explainable to an auditor, regulator, or affected user, or when a support team cannot answer why a specific prediction was made.

deepchecks-tests

Run Deepchecks suites (data integrity, train-test validation, model evaluation) on tabular / NLP / vision data + models. Pass `result.passed_conditions()` to CI to gate on regressions; the same checks run during research, CI, and production monitoring per the Deepchecks lifecycle posture. Use before training to catch train-test leakage and data-integrity defects in a tabular, NLP, or vision dataset, and to re-run the same suite on production samples to detect drift.

evidently-monitoring

Use Evidently OSS (100+ evaluation metrics, declarative testing API) to detect data drift, target drift, and model-performance regression, wired into CI as a gate (a Report run with include_tests) and into production monitoring as a continuous check; reports as HTML + JSON for both human review and pipeline assertions. Use when you need a drift or quality gate, or a scheduled monitoring job, for a tabular ML model. Built on the Evidently API specifically: for DeepChecks-based validation suites use deepchecks-tests instead.

fairlearn-fairness

Compute group fairness metrics (selection rate, demographic parity, equalized odds) per sensitive feature with `MetricFrame`, then mitigate disparities using Reductions algorithms (`ExponentiatedGradient` with constraint = `DemographicParity`/`EqualizedOdds`). Wire group-disaggregated assertions into the model-evaluation gate. Use when a model's decisions affect people and a stakeholder, auditor, or regulation (ECOA, GDPR Art. 22, EU AI Act high-risk) requires evidence of per-group outcomes, or when someone reports the model treats a specific group worse.

model-performance-regression-gate

Computes held-out metrics (accuracy, F1, AUC, RMSE) for a retrained model and compares them against the current production model, failing promotion when any metric regresses beyond a configured tolerance. Adds per-segment checks via Deepchecks WeakSegmentsPerformance so a model that improves globally but regresses on a key slice is still blocked. Use when a retrained model is a candidate for promotion and the CI pipeline must enforce a per-metric pass/fail gate before the artifact is pushed to the model registry.

model-risk-evidence-matrix

Assigns a machine learning model to a low, medium, or high risk tier from what its predictions decide about people, then derives the fairness and explainability evidence that tier must produce: group metrics per declared sensitive feature, intersectional breakdowns with per-cell counts, vulnerability scan categories, a drift monitoring plan, and per-prediction explanation logs. Supplies conventional demographic parity difference bands, a per-vulnerability-category blocking table, and evidence rules that mark a bundle incomplete or self-contradicting. Use when a model release candidate is up for promotion and someone must decide which fairness artifacts are mandatory rather than nice to have, or when a model card declares a risk tier and the attached evidence bundle has to be checked against what that tier demands.