As developers (or end users), we constantly have to switch contexts; juggling has become the standard in today's technology landscape; it’s just part of life. The fortunate thing is that we’ve been given a minion to take on our grunt work, and that minion’s name is Now Assist. Like most minions, they can’t be expected to read your mind without being taught how to do so. Lucky for us, we have  “Skill Kit”.

Now Assist Skill Kit is a way for us, as developers, to build AI-driven features for our organization to enable us to free up those much-needed cycles. I want to take a few minutes to present you with an overview of Skill Kit and some of what it has to offer.

If we navigate over to Now Assist Skill Kit > Home in our filter navigator, we will be presented with a  landing page from where we can select “Create Skill” to get started.

Give your skill a descriptive name and description to get started. For now, we will leave the Default provider and API. Later in this serie,s we will discuss connecting Now Assist to other AI providers and APIs. 

For this overview, we will be testing with the “Now LLM Generic” provider. Once we’ve named and saved our new skill, we can begin testing and configuring it.

We are greeted with a sample prompt to get familiar with how the inputs work and to test the output of our prompt.

Initially this screen will NOT have an input available to us but we can easily add it within the “Skill input” portion of the form.

Once we’ve selected our data type, in this example, we’re using the type “Record”, we will be prompted for additional information about the record type. This additional information will give context to the LLM when our skill runs. 

Here we’ve selected the incident table and chosen an incident to be used for testing. Now that we have our skill configured with an input we can configure our prompt to actually use the inputs we’ve created. By default, you are given an example prompt which we will use for testing.

Starting at the “Prompt” window we can see the context given from our input record. This input record uses variable substitution by wrapping your input information with double curly braces “{{ }}”. Within these curly braces we can dot walk to record fields as normal. 

1. Parameters given to the AI, this is general idea of what we want it to accomplish
2. Context, this is context from the record we are targeting that will give the AI additional details needed to accomplish its goal. 

From here you can make any additions or adjustments to the parameters or add additional context if needed. Once satisfied with your prompt, you can test the skill by selecting the “Run Test” button. This will run the prompt against your selected AI and return 2 things, a response and a grounded prompt. 

The response is what the user will see when using the prompt in the real world. The grounded response is simply your prompt but with variable substitution populated with record context. 

The last steps needed to deploy our new skill are to finalize the skill and figure out how we’re going to deploy it. First we need to Finalize the prompt by selecting the “Finalize Prompt” button, this will allow us to deploy and publish the skill. Once we’ve finalized we can jump to Skill Settings > Deployment Settings to select a deployment method. In this example we’re going to deploy as a ui action on the incident table.

Interested to learn more? Check out other posts in our Now Assist series, or reach out to chat@rapdev.io.