> ## Documentation Index
> Fetch the complete documentation index at: https://docs.oasm.io/llms.txt
> Use this file to discover all available pages before exploring further.

# Skills

> Give the agent specialized instructions it loads on demand, and manage your workspace's own skills

A **skill** is a named set of instructions in Markdown, such as a testing methodology or a step-by-step procedure, that the agent loads only when a task calls for it. Skills keep the agent's standing prompt short: the agent always sees each skill's name and description, and reads the full content only when it decides the skill applies.

Open them from **Settings → Skills** in the AI Studio sidebar. The header reads "Specialized instructions the agent can use to perform tasks".

![Agent settings Skills section with an Add Skill button and a list of skills marked SYSTEM, including api-testing, browser, command-execution, exploit-verification, network-service-testing, nuclei-template-generator, osint-and-assets, project-builder and reconnaissance, each with an expand arrow and an enabled switch](https://mintcdn.com/oasmplatform/JcAER2Ib37AN8_8A/images/screens/ai-skills.webp?fit=max\&auto=format\&n=JcAER2Ib37AN8_8A\&q=85\&s=403e65014a38b48991b072c576acb38e)

Settings → Skills: system skills are marked SYSTEM and are always on

## How the agent uses skills

1. Every conversation's system prompt lists the available skills: all system skills plus the workspace skills that are turned on, each as its name and description.
2. When your request matches a description, the agent calls its `load_skill` tool with the skill's name and receives the full content.
3. The agent follows those instructions for the rest of the task.

This works in both Ask mode and Agent mode. Subagents work differently: each one is offered only the skills its [agent definition](https://docs.oasm.io/ai-studio/settings/agents.md) names, and a named skill that was deleted or turned off is left out.

Because the agent picks skills by their description, write the description as "when to use this skill", not as a title.

## System skills

System skills ship with OASM. They are always on, marked **SYSTEM**, and cannot be edited, turned off or removed. Click a row to expand it and read the full content.

| Skill | What it covers |
| - | - |
| `api-testing` | API security testing: endpoint and schema discovery, object- and function-level authorization, mass assignment, injection through JSON or GraphQL, rate and resource abuse |
| `browser` | Driving a headless browser on the conversation's worker: JavaScript-heavy pages, logins, multiple identities, forms, cookies, storage, console and network logs |
| `command-execution` | Running security tools on the worker: built-in tools, connector tools, shell utilities, paging long output |
| `exploit-verification` | Turning a suspected weakness into a confirmed or refuted one with the least intrusive proof, a negative control and calibrated severity |
| `network-service-testing` | Non-destructive assessment of exposed non-web services such as TLS, SSH, mail, databases, caches, file shares and remote-access ports |
| `nuclei-template-generator` | Writing a valid Nuclei template from a description and saving it as a custom template of the workspace |
| `osint-and-assets` | Passive asset discovery: certificate transparency, ASN and netblock expansion, third-party services and infrastructure the target no longer controls |
| `project-builder` | Turning a scanning goal in plain words into an OASM [project](https://docs.oasm.io/project.md) with a configured tool pipeline and schedule |
| `reconnaissance` | Attack-surface mapping: subdomains, DNS, live hosts, ports, services, technologies, TLS, content discovery and parameters |
| `vulnerability-analysis` | Deep analysis of CVEs: CVSS, affected versions, exploit maturity and remediation |
| `web-app-testing` | Authorized, non-destructive web application testing aligned with OWASP: authentication, authorization, injection, XSS, SSRF, file handling, business logic |
| `web-research` | Researching CVEs, advisories and other security information that is not in the workspace |

## Add a workspace skill

Workspace skills are shared with every member of the workspace and are offered to everyone's agent as soon as they are saved.

**1. Open the Add Skill dialog**

In AI Studio, open **Settings → Skills** and click **Add Skill** in the header.

**2. Name the skill**

In **Name**, enter a short unique name, for example `code-review`. The agent loads the skill by this name.

**3. Describe when to use it**

In **Description**, write one line that tells the agent when the skill applies, for example "Use when reviewing a web app's login and session handling".

**4. Write the content**

In **Content**, write the instructions in Markdown: when to apply the skill, the steps to follow, what to check and what to hand back.

**5. Save**

Click **Add Skill**. The console confirms with "Skill created". New skills start turned on.

![Add Skill dialog with the subtitle Create a new skill for the agent, a Name field with the placeholder e.g. code-review, a Description field, a Content box for Markdown instructions, and Cancel and Add Skill buttons](https://mintcdn.com/oasmplatform/JcAER2Ib37AN8_8A/images/screens/ai-skill-add.webp?fit=max\&auto=format\&n=JcAER2Ib37AN8_8A\&q=85\&s=d483c03b1f4b5d80b65e1f7374fe8e1d)

The Add Skill dialog

### Fields and rules

| Field | Rules |
| - | - |
| **Name** | Required, up to 255 characters, and unique within the workspace. A duplicate fails with `Skill with name "…" already exists in this workspace` |
| **Description** | Required |
| **Content** | Required. Markdown |

> [!WARNING]
>
> Do not give a workspace skill the same name as a system skill. The save is accepted, but the agent always loads the system skill of that name, so your skill is never used.

## Turn a skill on or off

Use the switch on a workspace skill's row. A skill that is off shows a **Disabled** badge: it is left out of the agent's skill list and `load_skill` refuses it, for subagents too. Turn it back on to make it available again. System skills cannot be turned off.

## Edit or remove a skill

**1. Edit**

Click the pencil icon on a workspace skill's row. The **Edit Skill** dialog opens with the same fields. Click **Save Changes** ("Skill updated"). Renaming checks the new name for duplicates.

**2. Remove**

Click the trash icon, then confirm **Remove** in the **Remove Skill** dialog ("Skill removed"). Removing cannot be undone.

## Who can change skills

| Action | Requirement |
| - | - |
| See skills | `agent.read` |
| Add, edit, turn on or off, remove a workspace skill | `agent.write`, and you must be the **workspace owner** |

Other members see the **Add Skill** button and the controls, but the request fails ("Failed to create skill" and similar) because the API only accepts changes from the workspace owner. See [Permissions](https://docs.oasm.io/permissions.md).

## Related

> [card] [Agents](https://docs.oasm.io/ai-studio/settings/agents.md)
>
> Built-in agents and the skills each one loads

> [card] [Instruction](https://docs.oasm.io/ai-studio/settings/instruction.md)
>
> Standing instructions in every conversation


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.