Skills Overview

Skills are reusable browser-automation routines that your agent learns on its own. As an agent completes simulations, Marketrix distills the multi-step patterns it repeated — filling a form, signing in, searching and selecting a result, stepping through a wizard — into named Skills it can invoke as a single unit in future runs.

You don’t write Skills by hand. They are produced automatically from successful simulation runs and surfaced on the Skills page so you can review, inspect, and manage them.

What a Skill is

Each Skill captures one recurring workflow. A Skill has:

• Name — A short identifier for the workflow (for example, a login or search flow).
• Description — A one-line summary of what the Skill does.
• Version — Skills are versioned. When the agent learns a better way to perform the same workflow, it records a new version rather than overwriting the old one.

Skills are scoped to a single Application within your workspace. A Skill learned while exploring one Application is only offered to agents working on that same Application.

Why Skills matter

• Faster, more reliable runs — Instead of re-deriving a multi-step workflow from scratch every time, the agent invokes a Skill it already knows works.
• Captured product knowledge — Skills accumulate as a library of the workflows that matter in your Application, learned directly from how the agent navigated it.
• They improve over time — Each successful simulation can refine existing Skills or add new ones, so the agent’s repertoire grows the more it runs.

Where Skills come from

Skills are distilled from completed simulation runs. After a simulation task finishes successfully, Marketrix reviews the steps the agent took and extracts any reusable multi-step patterns as new Skill versions.

Single-step actions — a lone click, a single navigation — are never distilled into Skills. Those are already part of the agent’s built-in primitives, so there’s nothing new to learn from them.

Note

Skills are learned from simulation runs that complete successfully and contain enough steps to be worth capturing (by default, at least three). Short or unsuccessful runs don’t produce Skills.

Learn more

How Skills Work How Skills are distilled, versioned, and replayed.

Using Skills Review, inspect, and deprecate learned Skills.

Simulations The runs that Skills are learned from.

Agents The agents that learn and invoke Skills.