LM Studio Bionic Skills: The Complete Guide to Teaching Your AI Agent Repeatable Tasks (2026)

I've been glued to the LM Studio changelog since Bionic launched at the end of July, mostly because it's the one local AI feature I've seen actually change how I work day to day, rather than just being a cooler chat window. And a few weeks ago, they quietly shipped something that I think most people haven't fully clocked yet: Skills. If you've read my earlier Bionic setup guide, this is the natural next step — the feature that turns Bionic from "an agent that's good at tasks" into "an agent that gets better at your specific tasks over time."

Let's get into what Skills actually are, why they matter more than they sound like on paper, and exactly how to set one up.

What Is a Skill, Actually?

Think about the last time you did something multi-step with an AI agent and it went really well — say, you walked Bionic through reformatting a messy CSV into a clean report, or generating a specific style of commit message, or turning your rough notes into a particular document format you like. It worked. But the next time you needed that exact same thing, you had to explain the whole process again from scratch.

A Skill fixes that. It's a small file — a SKILL.md — that captures the instructions, steps, and any supporting reference material for that specific task, so Bionic can just run it the same way every time, without you re-explaining the process. It's built on an open standard called Agent Skills, which is the same format used by other agent tools like Claude Code and Codex. That last part matters more than it might seem — it means a skill isn't locked into one app.

Skill vs. System Prompt: What's the Actual Difference?

This trips people up, so let's be precise about it. A system prompt shapes how the model behaves across an entire conversation — its tone, its general priorities, its personality. A skill is narrower and more situational: it's a specific set of instructions for one particular repeatable job, and it only activates when it's relevant. You could have a general system prompt telling Bionic to "be concise and cite sources," and separately have five different skills for five different recurring tasks, each one only kicking in when that specific job comes up.

How to Create Your First Skill

There are two ways to build one, and both are far easier than writing a SKILL.md file by hand.

Method 1: Capture a Task You Just Completed

This is the one I'd start with. The next time you finish a multi-step task in Bionic that went well, just type:

"Create a skill based on what we just did."

Bionic will ask you a few clarifying questions — what should trigger this skill, what inputs it needs, any edge cases to handle — and then draft the SKILL.md file for you automatically. You're not writing documentation from scratch; you're reviewing and confirming what it already understood from doing the task with you once.

Method 2: Workshop a New Skill From Scratch

If you already know you want a skill for something you haven't done yet, describe the workflow directly: what inputs it takes, what output you expect, and any standards or formatting rules it should follow. Bionic will build the skill file collaboratively with you through the conversation, the same way you'd brief a new team member on a recurring task.

Using a Skill in Practice

Once a skill exists, there are two ways it gets used:

• Manually, with the @ shortcut. Type @ in the chat box to open the reference palette, then pick the skill by name. This is the reliable option when you want to be certain a specific skill runs.
• Automatically. Depending on the model you're using, Bionic can recognize when a task matches an existing skill and apply it without you asking — this depends on how well that particular model handles tool selection, so it's more consistent with stronger models.

Installing Skills Other People Made

You don't have to build everything yourself. If you find a skill someone's shared — say, in a GitHub repository — you can hand Bionic the URL and ask it to install the skill directly, or use the @Install Skill shortcut to walk through it. This is where the shared Agent Skills standard really pays off: a well-made skill for, say, structured meeting notes or a specific coding review pattern can be dropped in without you reverse-engineering how it works internally.

Bringing In Skills From Claude Code or Codex

If you've already built up a library of skills in Claude Code or Codex, you don't need to start over in Bionic. Head to Settings → Skills → Skills from other apps, and you can enable which existing skills should be visible to Bionic. You can also selectively disable individual ones from the ••• menu if you only want a subset available in a given context.

Practical Skills Worth Setting Up First

If you're not sure where to start, here are a few genuinely useful first skills based on common recurring work:

• A document formatting skill — turning rough notes into a consistently formatted report or summary in your preferred structure.
• A code review checklist skill — the same set of checks you always want run against a pull request or script.
• A research-to-outline skill — feeding Bionic a topic and having it always produce a consistent outline structure before writing.
• A data-cleaning skill — a fixed set of cleanup rules you always apply to messy spreadsheets or CSVs before analysis.

Start with something you already do more than once a week. That's the fastest way to feel the actual time savings.

A Few Things I've Learned Using Skills

• Be specific when Bionic asks clarifying questions during skill creation. A vague answer produces a vague skill that won't reliably trigger later.
• Name skills clearly. Since you'll often trigger them manually with @, a clear, specific name saves you from scrolling through a messy reference palette later.
• Review the generated SKILL.md occasionally. It's just a markdown file, so you can open and edit it directly if you want to tweak the instructions without recreating the whole skill.
• Don't over-build early on. Start with two or three skills for your most frequent tasks rather than trying to systemize everything on day one.

Why This Matters More Than It Sounds Like

Most "AI agent" features are about making one interaction better. Skills are about making every future interaction for a specific task faster than the last one, without you having to remember your own best practices each time. That compounding effect is genuinely the difference between "I used AI for this once" and "this is now just part of my actual workflow." Since this only just launched, most people haven't built any skills yet — which honestly makes now a pretty good time to get a small library going before it becomes standard practice for everyone.

If you set up a skill you're proud of, I'd genuinely like to hear about it — drop a comment with what it does. I'm collecting a few reader-submitted ones for a follow-up post.

Frequently Asked Questions

What are Skills in LM Studio Bionic?

Skills are reusable instruction files (SKILL.md) that teach Bionic a repeatable task, workflow, or body of knowledge, so you don't have to re-explain the same process every time you need it.

How do I create a Skill in LM Studio Bionic?

The easiest way is to complete a task with Bionic normally, then type a message like "Create a skill based on what we just did." Bionic will ask clarifying questions and generate the SKILL.md file for you.

Can I use skills I already made in Claude Code or Codex?

Yes. Bionic supports the same open Agent Skills standard used by tools like Claude Code and Codex, so existing SKILL.md files from those apps can be enabled for Bionic in Settings, under Skills from other apps.

How do I trigger a skill during a chat?

Type the @ symbol in the chat box to open the reference palette and select the skill by name. Depending on the model, Bionic can also detect and use a relevant skill automatically without being asked.

Can I install a skill someone else made?

Yes. Give Bionic a URL to a skill, such as a GitHub repository, and ask it to install the skill, or use the @Install Skill shortcut.

Is this the same thing as a system prompt?

No. A system prompt shapes the model's overall behavior for an entire conversation, while a skill is a specific, on-demand instruction set for one particular repeatable task, triggered only when relevant.

Do skills work with any model in LM Studio?

Skills work within Bionic, LM Studio's agent mode. Whether a model can automatically detect when to use a skill without being asked depends on that specific model's tool-use ability.

New to Bionic? Start with my Bionic setup and best models guide, then check out best system prompts for LM Studio and Ollama vs LM Studio to round out your local AI setup.