---
sourceDocument: Australia Enable AI
sourceDocumentLink: https://servicenow-prod.fluidtopics.net/r/pt-BR/intelligent-experiences

 Release :

    - australia

ft:locale :

    - pt-BR

ft:publication_title :

    - Australia Enable AI

ft:clusterId :

    - platai

bundleId :

    - platai

workflow :

    - Platform


---

# Scoping the skill

# Scoping the skill {#ariaid-title1}

* Versão de lançamento: Australia
* 
* Atualizado 12 de mar. de 2026
* 
* ![](https://www.servicenow.com/docs/portal-asset/ico-clock) 1 min. de leitura

Scoping the skill before you build it helps determine the requirements needed for the skill and the expected outcomes.

## Before you begin {#scoping-the-skill__section_b1c_nph_ccc}

Before you start to build a custom skill with Now Assist Skill Kit, you should first review the base system skills that are available in the Now Assist Admin console. Use these pre-existing skills whenever possible.

## User prerequisites {#scoping-the-skill__section_k1d_5ph_ccc}

Now Assist Skill Kit is meant to be used by a developer or someone with experience using generative AI. Because you must write the initial prompt template, you should:

1. Be knowledgeable about prompt engineering, including:
   * Having familiarity with the development and testing of machine learning systems.
   * Having informed expectations about the behavior and capabilities of a prompted large language model (LLM). You should understand:
     * The fundamentally probabilistic nature of LLMs.
     * That performance can vary greatly across different LLMs and different target tasks.
   * Having experience or training with writing and evaluating prompts for LLMs.
   {#scoping-the-skill__ul_wkb_gpj_ccc}
2. Have an in-depth understanding of the use case that they're trying to solve and the persona that they're building the skill for.
{#scoping-the-skill__ol_j54_bpj_ccc}

## Design {#scoping-the-skill__section_tln_vph_ccc}

Before you begin building a custom skill, you must think about the overall design and document the requirements. Consider the following:

1. What precisely do you want the skill to do?
   * What will be the inputs to the model?
   * What should be the outputs of the model? (Both content and format)
   * Who will be using the skill?
   * How can you characterize success?
2. What don't you want the model to do?
   * It's useful to think about the possible risks and downsides of using generative AI outputs so that you can try to guard against them during development and testing.
   * It's worth listing any specific model behaviors (LLM outputs) that would be detrimental or harmful in your specific use case.