Skills 101
Last updated: August 24, 2026
Skills are the working methods the AI Agent follows when you ask it for something. Where Agent Memory holds what the agent knows, a skill holds how to handle a given situation: which steps to take, in which order, and what to check.
Each skill has two halves. The first says when it should be used, in which situation and off which keywords. The second is the skill itself: a set of instructions, a small workflow, or just context that should be loaded whenever that situation comes up.
You will find them in Settings > Skills.
Topo skills
Every workspace ships with three built-in skills, ready to use:
Skill | Category | What it covers |
|---|---|---|
Lead search | Sourcing | Create, refine and run prospecting lead searches: translating your ICP into structured and AI filters, targeting |
Outbound messaging | Messaging | Draft cold email and LinkedIn sequence templates: copywriting, step count, cadence, follow-ups |
Sequence performance | Analysis | Interpret sequence metrics and benchmarks, diagnose underperforming sequences and suggest fixes |
Each card has an On toggle. Turn a skill off and the agent stops using it. Use the category chips (All, Sourcing, Messaging, Analysis) or the search box to filter.
The … menu on a card lets you edit, duplicate, archive or delete the skill. This applies to the three built-in Topo skills as well, so you can adapt any of them to how your team works. Topo does not rewrite your skills in the background: a skill changes when you edit it, or when you ask the AI Agent to update it.
Archiving a skill and turning its Status toggle off both stop the agent from using it, and both are reversible. Deleting a skill is different: it disappears immediately for everyone who could see it, with no undo.
The Your skills tab holds the skills you build yourself. It starts empty.
Creating your own skill
You can either ask the agent to create a skill (and describing it) or you can create it yourself.
Click + New skill and fill in the form:
Title (required). What the skill handles, for example "Handle a pricing question".
Identifier (required). Used by the agent to refer to the skill. Derived from the title, and editable.
Description (required). When the agent should use this skill, for example "Use this when a lead replies asking for pricing." This text is what drives discovery, so write it as a trigger condition rather than a summary of the content.
Scope. Toggle Everyone in the organization on to share it with the team, or leave it off to keep it to yourself. Scope cannot be changed after creation, so decide before you click Create.
Auto-discovery. On by default. The agent can use this skill on its own when it judges it relevant. Turn it off if you only want the skill used when you call it explicitly.
Status. On by default. Off means the agent will not use the skill at all.
Content (required). The instructions themselves. Say when to use the skill, then walk through the steps. The editor supports headings, bold, italic, bullet and numbered lists and tables, plus @ Insert reference to point at something that already exists in your workspace: lists, sequences, playbooks, lead searches and other skills.
Click Create skill to save it. There is no cap on the number of skills you can have, or on how long a skill's content can be.
How many skills can apply at once
At most three skills apply to a single message. If your request matches more than three, only three are used, and the agent tells you which ones it followed.
Two ways a skill gets used
The agent picks it. With Auto-discovery on, the agent reads the Description of every active skill and uses the one that fits your request. This is the default, and it is why the Description matters more than anything else on the form.
You force it. Type @ in the AI Agent chat and mention the skill by name. The agent uses that skill, no guessing involved. Handy when two skills are close, or when Auto-discovery is off.
Writing a skill that gets used
The Description does the work of getting the skill picked up. Name the situation and the words a user would actually type.
Keep one skill to one job. A skill that tries to cover pricing, objections and onboarding will fire at the wrong times.
In the Content, write the steps you would give a new joiner: order matters, and so does saying what to check before moving on.
Use it to load context as well as process, for example the rules behind one of your playbooks, or how your team qualifies an inbound reply.
Leave Auto-discovery on unless the skill is sensitive or narrow enough that you want to invoke it deliberately.
A skill changes how the agent works, never what it is allowed to do. It cannot widen the agent's scope or remove an approval step.
Still have a question? See 📄 FAQ: AI Agent, Memory, Skills & Playbooks