{"revision":"c10276e953456835cdbb1f3a60d58ae1980be353","document":{"id":24201,"path":"intelligent-experiences/now-assist-skill-kit/test-prompt-template.md","title":"Test a prompt","bundle":null,"doc_type":null,"release":"australia","body":"\n# Test a prompt\n\nAfter you create a prompt for your custom skill, test it before you finalize it to verify that it returns the expected results.\n\n## Before you begin\n\nRole required: sn\\_skill\\_builder.admin\n\n## Procedure\n\n1.  Navigate to **All** > **AI Skill Kit** > **Home**.\n\n2.  Select the skill that you created the prompt for.\n\n3.  In the Test prompt section, select **Run tests**.\n\n4.  Select a test record, such as an incident.\n\n5.  Select **Run test**.\n\n    **Important:**\n\n    Testing a skill consumes at least one Now Assist assist for the base prompt call.\n\n    If the skill has LLM-judged evaluation metrics attached on the **Deployment and skill settings** tab, each metric execution consumes additional assists. Metric execution is billed as a custom call at one assist per 1,000 output tokens. Script-based metrics \\(metrics with a script instead of a judge prompt\\) don't consume assists.\n\n    For information about adding evaluation metrics, see [Configure deployment and skill settings](https://raw.githubusercontent.com/ServiceNow/ServiceNowDocs/australia/markdown/australia/intelligent-experiences/now-assist-skill-kit/configure-skill-settings.md).\n\n    |Tab|Description|\n    |---|-----------|\n    |Response|Result that the large language model \\(LLM\\) returns for the prompt.|\n    |Grounded prompt|Prompt with the data from your skill inputs and tools filled in. Use this view to verify that your skill inputs and tools return the correct data.|\n\n6.  If the skill is deployed as a flow, turn off the **com.glide.oneapi.fdih.async.quick.mode** system property, and then turn on flow reporting.\n\n    This property controls whether flow execution details are generated when running flows, subflows, and actions from a custom skill. You can use flow execution details to test and troubleshoot your flow, subflow, or action. For more information about the system property, see [Workflow Studio flow system properties](https://raw.githubusercontent.com/ServiceNow/ServiceNowDocs/australia/markdown/build-workflows/flow-designer-system-properties.md).\n\n7.  Refine the prompt and repeat testing as needed.\n\n8.  Select the run test history icon \\[Omitted image \"icon-nask-test-history.png\"\\] Alt text: to see the results from your previous run tests.\n\n\n## What to do next\n\nAfter you test your prompt, finalize the prompt and publish the skill. To learn more about publishing a skill, see [Finalize and publish a skill](https://raw.githubusercontent.com/ServiceNow/ServiceNowDocs/australia/markdown/australia/intelligent-experiences/now-assist-skill-kit/publish-skill.md).\n\nIf you have not configured the deployment settings for your skill, see [Configure deployment and skill settings](https://raw.githubusercontent.com/ServiceNow/ServiceNowDocs/australia/markdown/australia/intelligent-experiences/now-assist-skill-kit/configure-skill-settings.md).\n\n**Parent Topic:**[Using AI Skill Kit](https://raw.githubusercontent.com/ServiceNow/ServiceNowDocs/australia/markdown/australia/intelligent-experiences/now-assist-skill-kit/using-now-assist-skill-kit.md)\n\n**Related topics**  \n\n\n[Create a skill]()\n\n[Create a prompt]()\n\n[Use prompt assistance]()\n\n[Evaluate a prompt]()\n\n[Finalize and publish a skill]()\n\n[Activate a skill]()\n\n[Call a custom skill from a script]()\n\n"}}