--- title: "Claude skills ยท Reusable agent skills for Claude Code and teams" description: "Agent skills package the instructions, scripts and reference files for one task. Install skills from the registry or write your own, and share them with your team." canonical: "https://inference.eu/skills/" language: "en" --- Agent skills registry # Claude skills and agent skills that teach your AI agents how your team does a task A skill is a folder with plain-language instructions, plus any scripts or reference files a task needs, such as preparing a monthly report or checking a contract. Agents load it when the task comes up, so they follow your way of working every time. [Contact sales](https://inference.eu/contact/)[How it works](#features) ## How it works - ### A registry to start from Browse ready-made skills for common work and install the ones that fit. - ### Write your own in plain language A skill is a SKILL.md file with instructions, plus the files it needs. Version it in Git like any other code. - ### Shared across your team Publish a skill to your workspace once, and every agent your team runs can use it. - ### An open format Skills follow the open agent skills format, so they also work in Claude Code, Codex and other agents. - ### Scripts and reference files Bundle the scripts, templates and reference documents a task needs, so the agent runs tested code instead of improvising. - ### Loaded only when needed Agents see a short description of each skill and read the full instructions only when a task calls for it, so many skills can sit side by side. ## Skills teams write first - ### Reporting Pull the same numbers from the same places and format them the way your management expects. - ### Contract and policy checks Review a document against your own checklist and flag what needs a lawyer. - ### House style Write emails, posts and documentation in your tone, with your terms and spellings. - ### Runbooks and routines Turn the steps for releases, incident handling or month-end close into skills that any agent can follow. ## Skills: common questions The questions buyers, developers and legal teams ask us most. Something missing? Ask us directly. A folder containing a SKILL.md file of instructions, plus any scripts or reference files the task needs. An agent reads the instructions when the task calls for it and runs the scripts as part of the work, so the result follows your method rather than the model's guess. Claude skills, which Anthropic calls Agent Skills, use this same folder format, with a SKILL.md file holding a name, a description and instructions, plus optional files. Anthropic published the format as an open standard, so a skill written for Claude also works with other agents that read it. Yes. Claude Code reads skills from a skills folder in your project or home directory. Install skills from the registry or your workspace there, and Claude Code uses them just like any skill you wrote yourself, alongside the agents on the platform. Yes. Skills use the open agent skills format, which Claude Code, Codex and other agents also read. You are never locked in, because a skill you write here can be copied into a repository and used anywhere the format is supported. Yes. Skills you write stay in your workspace unless you choose to publish them. Your team and the agents you allow can use them, and nothing reaches the public registry without a deliberate decision on your side. Start with one task your team repeats. Describe when the skill applies in a single sentence, then write the steps as you would explain them to a new colleague, with examples of good output. Move long reference material into separate files and add scripts for anything that must be exact. An MCP server gives an agent access to an app or system, such as your calendar or accounting software. A skill tells the agent how to do a task, such as preparing the board report from those numbers. Useful agents combine both, with connections for access and skills for method. A skill can contain scripts, so read it before you install it, as you would any code from the internet. Check what the scripts do and which tools the instructions ask for. Many teams copy a registry skill into their own workspace, review and adjust it, then let agents use it. Each skill starts with a short name and description. Agents keep those descriptions in view and load the full instructions only when a task matches, which keeps the context small. A clear, specific description is therefore the most important line in any skill. Yes. A skill is a folder of plain files, so you can keep it in Git, review changes in pull requests and roll back a version that made results worse. Publishing to your workspace then makes the agreed version available to every agent your team runs. Yes. A skill is instructions and files, not model-specific code, so agents on the platform use skills with any of its open-weight models. Larger models follow long, multi-step skills more reliably, so test a skill on the model you plan to run it with. Publish it to your workspace. Every agent your team runs can then use it, and colleagues can install it in Claude Code or another agent. When you publish an update, the whole team works from the new version. [Ask a question](https://inference.eu/contact/) ## Move your AI workloads to EU infrastructure Tell us which models you use and roughly how much traffic you have. We reply within one working day with pricing and next steps. - Reply within one working day - Pricing for your volume - DPA ready to sign