What Bionic gained with skills support
A skill packages a capability or a body of knowledge — or just a job you keep repeating — into a shortcut you can call by name. The wrapper is Agent Skills, an industry standard whose concrete form is a SKILL.md file. Because Bionic takes that format as-is, nothing about how you write skills has to change.
Three verbs: use, install, create
The release adds three operations.
| What you want | How it is done |
|---|---|
| Use one | Type @ in the composer and pick from the palette |
| Install one | Hand Bionic a URL and ask, or call @Install Skill |
| Create one | Say so mid-conversation: make this into a skill |
The skills you already hold — the ones sitting there for Codex or Claude Code — are usable as they are. Nothing has to be rebuilt for this app; the point of the standard is that the library moves with you.
"Bionic now supports using, installing, and creating agent skills."／"Skills are a way to capture specific knowledge, a capability, or a repeatable task in a sort of shortcut—using an industry standard called Agent Skills (SKILL.md files)."／"The Bionic agent can use skills in the standard format: install skills from the web, create skills from a conversation, and use skills you already have in other apps like Codex and Claude." (intro) — from the LM Studio blogView official source →
Reusing the skills you already have
Which of those skills Bionic can see is under your control. The path is Settings, then Skills, then Skills from other apps. Anything you would rather it ignored gets switched off from the ••• menu.
If you already have a SKILL.md sitting on disk, there is less to it than that. Drop the file into Settings > Skills and the agent can use it.
"We shipped support for using skills you already have in other apps. Head to Settings > Skills > Skills from other apps and manage which skills you want to be visible to Bionic. You can disable certain skills in the ••• menu."／"Alternatively, if you have a SKILL.md file ready, you can just drop it in Settings > Skills and make it available to your agent to use." (Use skills from other apps / Install Skills) — from the LM Studio blogView official source →
Calling a skill and writing one
Both operations happen in the composer, and neither takes long to explain.
The @ syntax, and when Bionic decides for itself
Typing @ opens the reference palette for files and skills; pick one from there. Referencing a skill in chat makes Bionic open the file and work out what to do with it.
With some models Bionic may be able to reach for the skill itself, without you referencing anything. The post is careful here, saying only that this depends on the model, without naming which models manage it.
"The simplest way to use a skill is to type the @ character, which brings up the file or skill reference palette. When you reference skills in the chat, Bionic will inspect the skill file and decide how to act."／"Once the skill is created, you will be able to invoke it with the @ syntax. Depending on the model you use, Bionic might be able to automatically use a skill without you even asking." (Use Skills / Create Skills) — from the LM Studio blogView official source →
Creating — and installing — from a conversation
LM Studio describes two. One starts from work that is already finished; the other builds something new alongside the agent.
The first has an obvious moment. You and Bionic just finished a job and the method held up. Next month you would rather not rebuild it from memory. Say so — make a skill out of what we just did — and Bionic asks its clarifying questions before writing it. From then on it answers to @.
Installing works the same way. Give the agent a URL and let it do the work, or call @Install Skill. Placing the file yourself is possible, but it is not the route the post recommends.
"Bionic can create skills for you—just ask! Consider a couple of patterns: one is creating a skill based on a task you just completed. Another is workshopping and creating a skill together with Bionic for something new."／"Similar to creating skills, the best way to install skills is by asking Bionic to do it for you. Give the agent a URL and ask it to install the skill, or invoke @Install Skill ." (Create Skills / Install Skills) — from the LM Studio blogView official source →
Where the skills run: local, or cloud
What runs underneath decides which one you reach for. Bionic is the agent app for open models that LM Studio released on July 16, 2026. It ships separately from LM Studio itself, and the older app remains the place for low-level configuration.
Local models on the LM Studio runtime
Current local LLMs — models that run on your own hardware — download from inside Bionic and serve both ordinary chat and agentic work. The LM Studio runtime executes them. For jobs too heavy for the machine in front of you, LM Studio Secure Cloud switches you to a large open model, and requests sent that way are handled transiently rather than kept. Voice input never makes that trip at all: transcription for the voice keyboard finishes on the device.
For coding specifically, the models LM Studio names are GLM 5.2 and Kimi K2.7 Code. How far a local model actually gets you on coding work is measured in our test of local LLM coding performance.
"Download the latest local LLMs directly within the Bionic app, then use them for simple chats or advanced agentic tasks. Local models in Bionic are powered by the LM Studio runtime."／"Bionic supports the latest frontier open models for your most complex tasks, running on the LM Studio Secure Cloud." ／ "When using cloud models, your requests are processed transiently and are not retained after the request completes."／"Bionic works with powerful open models like GLM 5.2 and Kimi K2.7 Code, so you can build more while keeping costs under control."／"Bionic is a new, separate app from LM Studio. For advanced low-level configuration, you can continue to use LM Studio alongside Bionic."／"You can use local models or switch to open-source models in the cloud for heavier tasks, all while staying in control of your privacy and AI spend."／"Use Bionic's voice keyboard with local transcription to speak through ideas, prompts, and edits - all entirely locally on your device, using state-of-the-art local audio models." (Natively Local / Cloud inference with Zero Data Retention by default / Bionic for Coding / Getting started / intro / Offline voice transcription) — from the LM Studio blogView official source →
Read a shared skill before you trust it
A skill is a set of instructions that will be executed, so where the file came from is a live question. Docker has written about setups that share one skills store across several sandboxes: edit a skill inside one and it becomes input to the next sandbox that loads it. The same post draws the boundary, though. Docker calls that store sandbox state — an edited skill will not start itself on your own machine, and the exposure stays inside that set of sandboxes.
Open a distributed SKILL.md and read it before installing. The more widely a standard format spreads, the more that habit earns. Attacks that quietly redefine an already-approved path are covered in the approved command that ran someone else's code.
"Every sandbox sharing that store sits inside one trust boundary, so a skill modified inside one becomes an input to the next that loads it. The store is sandbox state though, and a modified skill does not by itself execute on your host, so the risk runs sandbox to sandbox rather than sandbox to host." (How Docker Sandboxes Contain This at the Execution Layer) — from Docker's official blogView official source →
Published skills are often documented as web pages rather than files. Convert the page HTML to markdown; the headings and the step order come through intact. That output is a usable first draft of a SKILL.md.
Getting the most out of LM Studio skills
Three things carry the rest. The format is standard SKILL.md, so a skill written for another agent is already compatible. Calling is @; installing and creating are requests, not file operations. And the skills you have collected elsewhere get exposed to Bionic one by one from Settings — a checklist, not a migration. Stay on local models and the LM Studio runtime does the work. Voice transcription never leaves the device either. What you end up with is a library grown in other apps, running on hardware you own.
If you are starting from nothing, the order that wastes the least time is to expose what you already have first, work normally until you notice what repeats, and only then ask Bionic to capture those sequences. The ones written afterwards come out of a session that already succeeded, which is why the create-from-conversation path exists at all. Read anything you install from outside before it goes in, and keep the switch list under Settings honest — a skill nobody uses is still a set of instructions the agent may decide to read.
