Introducing the Skill Manager and Skill Repository

Dominique Rolink

Install trusted skills once, then use them across agents and projects.

Two new features have just arrived in AI Assistant to address this issue: the skills manager and the skill repository. Together, they make skills easier to discover, trust, and reuse. Instead of keeping skills tied to one agent or project you can install them once and use them wherever they are needed while managing them from inside the IDE.

In practice, a skill gives an agent reusable capabilities for a specific task. The problem is that skills often stay tied to a single setup. They live with one agent, one repository, or one machine. Skills Manager changes that by making them reusable across projects and supported agents.For a quick introduction to what skills are, check out this great explanation from PyCharm Content Creator Kristel Cocoli.

Skills manager: One place to install, manage, and reuse skills

Skills are useful because they give agents reusable capabilities for specific developer tasks, whether that is debugging CI failures, working through PR comments, automating browser flows, or converting Java to Kotlin. The problem is that they often stay locked to a single setup.

Skills Manager fixes that by adding a new IDE layer for skill management. That means developers can install skills once inside the IDE and make them available across supported agents and across all projects opened in that IDE, instead of rebuilding the same setup over and over.

It also supports different ways of working depending on the task. Some skills belong at the IDE level, where they stay available across projects for an individual developer. Some belong at the project level, where they can travel with the repository and be shared through version control. Others are best kept agent-specific, tied to a dedicated workflow like CI triage, frontend work, or code review.

That is the core improvement: Skills Manager gives developers one place to discover recommended skills, choose the right scope for each one, and keep those skills available where the work actually happens.

Skill repository: A verified starting point

The skills manager makes skills easier to use. The skill repository makes them easier to get started with.

At launch, the repository gives you a JetBrains-filtered and verified list of skills, organized for easier discovery and reuse. Instead of building a collection from scratch or managing it by hand, you get a curated starting point with skills that are ready to be installed.

The repository is also designed to make adoption safer. New additions are screened to detect prompt injection, data exfiltration, and malicious code patterns. Attribution is preserved by using the skill’s own author metadata when available, or otherwise crediting the upstream maintainer or organization. That gives you a practical starting point you can trust, with useful skills, a safer adoption path, and clear credit to the people who created them.

Recommended skills to try first

Here are just a few examples from the repository that show the range of tasks where skills can be used to guide agents with improved accuracy. To view the full repository, check out this link: https://github.com/JetBrains/skills. 

• React-best-practices –  Reusable React and Next.js guidance for writing, reviewing, and refactoring frontend code.
• postgres-best-practices – Practical guidance for Postgres queries, schema design, performance, and security.
• playwright – A structured way to automate and debug real browser flows.
• pnpm – Better support for pnpm-based JavaScript projects, including workspaces and CI usage.
• kotlin-tooling-java-to-kotlin – Support for disciplined migration from Java to idiomatic Kotlin.

Transparency and limitations

Skills Manager introduces an IDE-wide layer so developers can install skills once and make them available across supported agents and all projects opened within the IDE. That is the recommended default experience, but support is not universal yet.

• Today, IDE-wide skill storage is supported in AI Assistant Chat for Codex and Claude Agent. Support for Junie and other ACP agents is coming.
• For CLI workflows, the experience is different. CLI agents cannot use IDE-installed skills, so terminal-based workflows still rely on project-level or agent-specific installation.
• Support also differs slightly between agents. For example, most agents can work with shared project-level skills, while Claude Agent uses its own agent-specific location  (.claude folder), instead of the shared project location.
• The repository will also keep expanding over time, including with more IDE-specific skills.

Get started

To get started, install the AI Assistant plugin in your JetBrains IDE. In the AI chat, click the + button and go to Skills to add, remove, and manage the skills that fit your workflow.