> For the complete documentation index, see [llms.txt](https://docs.karini.ai/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.karini.ai/skills-library/skill-integration-with-prompts.md).
# Skill Integration with Prompts
Skills enhance agent capabilities by injecting domain-specific expertise into prompt contexts. This section covers how to integrate both **native** (platform-provided) and **custom** (organization-created) skills with your prompts.
### Prerequisites
| Skill Type | Requirement |
| ----------------- | -------------------------------------------------------------- |
| **Native Skills** | No prerequisites — always available |
| **Custom Skills** | Skill must have at least one published version (V1, V2, V3...) |
> **Note:** Draft-only skills are not visible in prompt. Publish the skill in the Skills Library before attempting to use it in a prompt.
### Accessing the Skills Interface
1. Navigate to **Prompts** in the sidebar.
2. Open an existing prompt or create a new one.
3. Click the **Skills** tab in the prompt editor.
The Skills tab displays two sections:
* **Native Skills** : Platform-provided skills, always available.
* **Custom Skills :** Skills created within your organization.
### Native Skills
Native skills are organized by category and provide general best practices maintained by the platform.
#### Browsing Native Skills
* Scroll to the **Native Skills** section in the Skills tab.
* Skills are grouped by domain (e.g., Security, Documentation, Testing).
* Use the search bar to filter by keyword.
* Click any skill name to preview its content.
#### Viewing Skill Details
Click a skill name or its info icon to open the detail view, which includes:
* **Description:** What the skill does.
* **SKILL.md** : Core instructions the agent will follow.
* **Supporting Files:** Templates, examples, and references.
Review these to confirm the skill's guidance aligns with your prompt's intent.
#### Adding a Native Skill to a Prompt
1. Locate the desired skill in the **Native Skills** section.
2. Verify the description matches your use case.
3. Click **Add** next to the skill name.
4. The skill appears in the **Active Skills** list at the top.
5. Click **Save** or **Publish** to persist the change.
### Custom Skills
Custom skills are organization-specific and support full version control.
#### Browsing Custom Skills
* Scroll to the **Custom Skills** section in the Skills tab.
* Use the filter tabs to narrow results:
| Tab | Shows |
| ------------ | -------------------------------------- |
| **All** | Every custom skill in the organization |
| **Active** | Currently enabled skills |
| **Inactive** | Deactivated skills |
* Skills display their current version (e.g., V2, V3).
#### Viewing Skill Details
Click a skill name or its view icon to see:
* **Current Version** : The version pinned to this prompt.
* **Latest Version** : The most recently published version.
* **Description** : Skill purpose and scope.
* **SKILL.md Content :** Instructions the agent will follow.
* **File Structure :** Supporting files and templates.
#### Updating a Skill Version
Custom skill versions can be updated without detaching the skill:
1. Open the skill detail view.
2. Select the desired version from the **Version** dropdown.
3. Click **Update**.
4. Republish the prompt to apply the change.
### Removing a Skill
1. Locate the skill in the **Active Skills** list.
2. Click **Remove** or the trash icon.
3. The skill returns to the available skills section.
4. Click **Save** to persist the change.
Once removed, the skill's instructions are no longer injected into the agent's context during invocations.
### Publishing and Testing
After saving a prompt with skills attached:
1. **Test** the prompt to verify skills are invoked correctly.
2. **Publish** the prompt to make skills active in recipes and copilots.
### How Agents Use Skills
Agents discover and execute skills through a three-stage process.
#### Stage 1 - Discovery
The agent's system prompt includes a catalog of available skills with names and descriptions. The agent reads this catalog to understand what capabilities are available.
**Example catalog:**
```
Available Skills:
- salesforce-sync: Synchronize customer data with Salesforce CRM
- pdf-generator: Generate professional PDF documents from structured content
- data-transformer: Transform CSV/JSON data between formats with validation
```
#### Stage 2 - Selection
When a user makes a request, the agent matches the intent to the most appropriate skill based on semantic alignment.
**Example:**
```
User: "Create a PDF report from this data"
Agent: Requires PDF generation → selects pdf-generator
```
#### Stage 3 - Execution
The agent reads the skill's detailed instructions, supplies the required parameters, and executes the skill. The result is incorporated into the agent's response.
**Example:**
```
Agent invokes: pdf-generator { content: "...", title: "Q4 Report" }
Skill returns: download_link: "https://..."
Agent responds: "I've generated your PDF report: [download link]"
```
