---
title: Using Skills
description: "What skills are from a user's perspective — packaged capabilities the agent draws on, how they shape clarifying questions, and how correction works."
---
> **For AI agents:** the complete documentation index is at [llms.txt](/llms.txt). Append `.md` to any page URL for its markdown version.
A **skill** is a packaged capability — a recipe that tells Matrix how to handle a class of request (writing a plan, reviewing code, deploying a contract). You don't invoke skills directly; the agent selects the right one based on what you ask.
## How skills shape the conversation
- **Clarifying questions.** A skill knows what it needs. If a required detail is missing, Matrix asks one focused question rather than guessing.
- **Typed structure.** Skills extract a typed plan from your prose, so your intent survives multi-step execution instead of drifting.
- **Honest gaps.** If something is genuinely unknown, Matrix marks it as a blocking gap and tells you, instead of inventing an answer.
## Correcting the agent
Because the plan is typed and inspectable, you can correct it structurally — adjust a constraint, change a target, raise a budget — without rewriting your whole request from scratch.
You can ask Matrix what it's about to do before it does anything irreversible. The rigorous rail always produces a reviewable plan for monetary or on-chain work.
The ideas behind skills, intents, and plans.