---
title: Understanding & Managing Agent Skills
description: What Agent Skills are and how to run them in NetworkOS — catalogue skills like /research and /extract-records, plus Managed skills built for your program.
---
[Skip to content](https://help-networkos.collaboration.ai/understanding-managing-agent-skills#main-content)
English
Show submenu for translations
/NetworkOS%20Word%20Mark/networkos-full-brand-800px.png?width=350&height=52&name=networkos-full-brand-800px.png)
Open main navigation
Close main navigation
- English
Show submenu for translations
- [Submit Feedback](https://help-networkos.collaboration.ai/submit-feedback)
[Submit Feedback](https://help-networkos.collaboration.ai/submit-feedback?hsLang=en)
This is a search field with an auto-suggest feature attached.
- There are no suggestions because the search field is empty.
1. [NetworkOS](https://help-networkos.collaboration.ai/?hsLang=en)
2. [Agent Skills](https://help-networkos.collaboration.ai/agent-skills?hsLang=en)
# Understanding & Managing Agent Skills
## Configured, repeatable capabilities that turn a request into sharp, actionable results.
### **What are Agent Skills**
Your Workspace is already conversational and data-aware from the moment it's set up — ask a question, get a data-backed answer, no Agent Skill required. **Agent Skills** are what take that from reactive to proactive.
An Agent Skill is a discrete capability that performs a specific platform task — a chained, configured workflow built around a defined process, rather than a one-off question. Instead of asking something and seeing what comes back, you're running a defined process that fetches the right data, assesses it against what matters to the task, and returns a result in the same shape every time.
Agent Skills can also be **chained** together, so a single request moves through multiple steps in sequence instead of running each one on its own.
There are two kinds available in your Workspace — catalogue and managed — covered below.
### **Catalog vs. Managed Agent Skills**
#### **Catalog Agent Skills**
Built once, maintained by Collaboration.Ai, and available in every Workspace by default. Catalogue skills are shared and global — one configuration applies everywhere, with no setup required on your end.
### **Catalog Agent Skills**
| **Agent Skill** | **What it does** |
| --- | --- |
| Research `/research` | combines your Workspace context with external sources to build a report or briefing |
| Summarize `/summarize` | generates a summary of a person, organization, or file |
| Search `/search` | semantic search across everything your Workspace knows |
**Managed Agent Skills**
A Managed Agent Skill is one Collaboration.Ai builds and configures for a single Workspace, scoped to a need unique to that program.
A Managed Agent Skill usually gets its own name once it's built — for example, one that matches an organization's profile against a specific category of funding, or one that surfaces a specific kind of internal reporting. If your Workspace has one, you'll see it by that name in the command menu.
**Note:** building a new Managed Agent Skill isn't self-serve — that's a conversation with your Collaboration.Ai team, who scope and build it around your actual workflow.
### **Running an Agent Skill**
**Where to find it:** the slash command menu in Terminal — type `/` to see every Agent Skill available to your Workspace.
Every Agent Skill has a Discovery Mode that determines how it runs: **Command** skills run when you call them from the command menu; **Ambient** skills can also join in automatically when they're relevant.
All four Catalog Agent Skills are Command-only — they act when you ask, not before. You can see each skill's Discovery Mode on its detail page in Workspace Settings.
| **Discovery Mode** | **Behavior** |
| --- | --- |
| Command | The skill is only available to the model when you explicitly type its slash command. |
| Ambient | The skill is available to the model automatically, based on conversation context — in addition to still being invokable by command. |
Ambient mode adds automatic invocation on top of manual access; it doesn't take manual access away.
For the mechanics of the slash command menu itself — opening it, what else lives there — see [*Slash commands*](https://help-networkos.collaboration.ai/the-terminal#slash-commands).
Enabling, disabling, and configuring Agent Skills for your Workspace is admin work. See [*Managing Agent Skills*](https://help-networkos.collaboration.ai/workspace-settings-overview#managing-agent-skills) in Workspace Settings.
- [Getting Started](https://help-networkos.collaboration.ai/getting-started?hsLang=en)
- [Terminal](https://help-networkos.collaboration.ai/terminal?hsLang=en)
- [Agent Skills](https://help-networkos.collaboration.ai/agent-skills?hsLang=en)
- [Artifacts](https://help-networkos.collaboration.ai/artifacts?hsLang=en)
- [Missions](https://help-networkos.collaboration.ai/missions?hsLang=en)
- [The Knowledge Layer](https://help-networkos.collaboration.ai/the-knowledge-layer?hsLang=en#main-content)
- [Your Data](https://help-networkos.collaboration.ai/the-knowledge-layer?hsLang=en#your-data)
- [External Data](https://help-networkos.collaboration.ai/the-knowledge-layer?hsLang=en#external-data)
- [Workspace Settings](https://help-networkos.collaboration.ai/workspace-settings?hsLang=en)
- [Trust & Safety](https://help-networkos.collaboration.ai/trust-safety?hsLang=en)
- [Release Notes](https://help-networkos.collaboration.ai/release-notes?hsLang=en)
- [Support](https://help-networkos.collaboration.ai/support?hsLang=en)
[/NetworkOS%20Logo%20Mark/networkos-logomark-black-512px.png?width=40&height=40&name=networkos-logomark-black-512px.png "networkos-logomark-black-512px")](http://collaboration.ai)
Copyright © 2026, Collaboration.Ai