---

name: validations-and-results description: "Create and run GX ValidationDefinitions, pass batch and suite parameters, choose result formats, interpret validation results, and retrieve unexpected rows." disable-model-invocation: true metadata: disco-role: operating license: Apache 2.0

Validations and Results

Use this sub-skill when a task needs to execute Great Expectations validations and interpret their outputs. It assumes the agent already has a GX context, a data asset or batch definition, and an expectation suite.

Route by task

• Read Validation definition workflows to create, save, retrieve, and run gx.ValidationDefinition objects, pass batch_parameters and expectation_parameters, or use the batch.validate() shortcut.
• Read Result formats and outputs before selecting ResultFormat.BOOLEAN_ONLY, BASIC, SUMMARY, COMPLETE, or a result-format dict, and when interpreting success, statistics, results, meta, and exception_info.
• Read Unexpected rows when a task asks for all failing rows, unexpected_rows, unexpected_index_query, caps, or backend-specific limitations.
• Read Troubleshooting to diagnose stale validation definitions, missing batch parameter keys, unresolved suite parameters, overlarge results, hidden exceptions, or result-format confusion.
• Run scripts/run_validation_smoke.py to verify the installed GX Python API with a tiny local dataframe validation that prints success and statistics.

Prerequisite routes

• For contexts, stores, Cloud/file/ephemeral mode, and project configuration, read ../contexts-and-configuration/SKILL.md.
• For datasource, data asset, batch definition, and batch-parameter setup, read ../datasources-and-assets/SKILL.md.
• For expectation and suite authoring details, row conditions, custom expectations, and $PARAMETER usage inside expectations, read ../expectations-and-suites/SKILL.md.
• For checkpoint orchestration, actions, notifications, and Data Docs updates, read ../checkpoints-actions-and-data-docs/SKILL.md.

Core public APIs

• Import GX as import great_expectations as gx.
• Construct validations with gx.ValidationDefinition(name=..., data=batch_definition, suite=expectation_suite).
• Persist and retrieve validations through context.validation_definitions.add(...), add_or_update(...), get(name), all(), and delete(name).
• Execute validations with validation_definition.run(batch_parameters=..., expectation_parameters=..., result_format=..., run_id=...).
• Validate one expectation directly against an already materialized batch with batch.validate(expectation, result_format=..., expectation_parameters=...).
• Use gx.ResultFormat enum values or equivalent strings/dicts for result verbosity.

Guardrails

• Do not use checkpoints for simple validation execution unless the task needs actions, notifications, or Data Docs.
• Do not assume every backend supports every diagnostic output; unexpected row retrieval and index-query output vary across Pandas, SQL, and other execution engines.
• Avoid COMPLETE and include_unexpected_rows on large batches unless the user explicitly needs row-level diagnostics and accepts the performance and memory cost.
• Keep validation code free of credentials and external services unless routed through the context/datasource sub-skills.