> For the complete documentation index, see [llms.txt](https://en.help.firstline.cc/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://en.help.firstline.cc/feature/ai-skill/use-ai-skills-in-ai-assistant.md).

# Use AI skills in AI Assistant

Explains how to apply AI Skills in AI Assistant, and how skills affect the capabilities, workflows, and data permissions available in that session.

AI skills allow the AI assistant to apply enterprise preconfigured working methods and capability scope in different work scenarios.

For example, an administrator can create a “customer complaint case consolidation” skill, requiring the AI assistant to organize issues only according to the customer, conversation, and ticket information actually found, without inferring missing content on its own; they can also create a “knowledge lookup” skill, allowing the AI assistant to focus on answering internal questions using enterprise knowledge.

After using a skill,**the AI assistant will apply that skill’s instructions on top of the original rules and permission restrictions, and determine which capabilities can be used this time according to the skill settings.**

## What does the skill affect?

When using skills in the AI assistant, it mainly affects two parts:

### Working method

Skill instructions can tell the AI assistant how this task should be handled, for example:

* Which key points need to be organized.
* What format the result should be presented in.
* Which information should not be inferred on its own.
* Which enterprise rules or wording should be followed when answering.

For example, the “customer complaint case consolidation” skill can require the AI assistant to consistently organize the results into four sections: “issue summary, known facts, items to be confirmed, and next steps.”

### Capabilities available this time

Administrators can also limit which AI assistant capabilities a skill can use.

This means that the same AI assistant, when no skill is applied or when different skills are applied,**the scope of work it can perform this time may be different**.

For example:

* “Customer complaint case consolidation” can allow queries for customer, conversation history, and ticket-related information.
* “Knowledge lookup” can keep only enterprise knowledge queries, preventing the task from extending to other data.
* “Customer service text editing” can focus on rewriting, tone adjustment, or translation, without also opening up data query capabilities.

{% hint style="info" %}
The capability settings in a skill are used to**limit and focus this task**, not to grant users new system permissions.
{% endhint %}

## How to use

When the enterprise has created skills available for the AI assistant and the current role has permission to use them, you can select the needed skill in the AI assistant and then enter the problem to be handled this time.

For example, after selecting “customer complaint case consolidation,” you can ask:

> Help me organize this customer’s most recent complaint issues, current handling progress, and items that still need confirmation.

The AI assistant will handle it according to the working method defined by the skill and will use only the capabilities allowed by that skill.

If you switch to another skill next time, or do not apply a skill, the AI assistant will again determine the capabilities it can use based on that time’s settings. A skill does not permanently change the functional scope of all subsequent conversations.

## Why limit the capabilities a skill can use?

More capabilities are not necessarily better.

Keeping only the capabilities truly needed to complete the task in a skill usually has the following advantages:

* **More focused tasks**: reduces the likelihood that the AI will extend the issue into unrelated directions.
* **More consistent results**: similar tasks will use more fixed information sources and processing methods.
* **Easier for users to understand the scope**: users know what this skill can mainly handle and what is outside its scope.
* **Reduce unnecessary queries**: avoid using too many data sources or capabilities for simple tasks.

Therefore, when creating a skill, it is recommended to start from the actual workflow and select only the capabilities necessary to complete that task.

## Skills do not break through the original data permissions

Even if a skill allows querying customers, tickets, knowledge, or other enterprise data, the AI assistant will still only obtain content that the current user is originally authorized to view.

For example, if a specialist originally cannot view tickets from a specific department, applying a skill that includes ticket query capabilities will not grant access to that department’s data.

{% hint style="warning" %}
AI skills do not elevate role permissions, nor can they bypass FIRST LINE’s original data visibility scope, security mechanisms, or other usage restrictions.
{% endhint %}

## Usage tips

When creating skills for the AI assistant, you can confirm the following in order:

1. **What kind of work does this skill need to complete?**
2. **What work rules should the AI follow?**
3. **What data or capabilities are needed to complete the work?**
4. **Which capabilities are actually not needed and should be turned off?**
5. **Which roles really need to use this skill?**

For example, “customer complaint case consolidation” mainly needs customer context, conversation records, and ticket information, so there is no need to open everything just because the AI assistant has other capabilities as well.

The value of a skill is not to let the AI assistant do more things at once, but to let the team fix**the handling method and necessary capabilities for a specific task**, making similar tasks easier to repeat and manage.