Distribute AI assistant skills and commands across workspaces with Lola

Use Lola to distribute team-authored skills, commands, and agents from a central Git repository into Che workspaces. Lola supports multiple AI assistants, including Claude Code, Cursor, Copilot, Gemini CLI, and OpenCode. It translates module content into each assistant’s native file structure, so developers receive a consistent set of standards and tools without manual setup.

A team maintains a Lola module in a Git repository containing skills (coding standards, guardrails), commands (on-demand audit macros), and agent definitions. Developers add a .lola-req file to their project referencing the module. Running lola sync fetches the module and installs it into the target assistant’s native directory structure.

Lola manages skills, commands, agents, instructions, and MCP server declarations. It does not persist assistant-specific settings, credentials, or plugin caches. For full Claude Code configuration persistence, see Additional resources.

Prerequisites
  • An AI assistant (Claude Code, Cursor, Copilot, Gemini CLI, or OpenCode) installed in the workspace container image.

  • Lola installed in the workspace container image or available on PATH. To install Lola:

    $ uv tool install "git+https://github.com/LobsterTrap/lola@53b3687975b373c856fc67b9a8f18c395c76fca4"

    The released package, uv tool install lola-ai, predates the agent-plugin layout and installs only the skills of such a module.

  • A Git repository containing a Lola module. At minimum, the module should contain:

    plugin.json
    skills/
      <skill-name>/
        SKILL.md
    dev.getlola/
      AGENTS.md
      commands/
        <command-name>.md
  • Git credentials configured in the workspace if the module repository is private. See How Git authentication works in cloud development environments.

Procedure
  1. Create a .lola-req file at the root of your project repository with a reference to the module:

    <git_repository_url>.git@main#assistant=<assistant> (1) (2)
    1 Replace <git_repository_url> with the URL of your Lola module repository. Replace <assistant> with the target assistant: claude-code, cursor, copilot-vscode, copilot-cli, gemini-cli, or opencode. Add subdirectory=<path> to the fragment when the module is not at the repository root.
    2 Keep the .git suffix: Lola reads the @main ref only from a clone URL that ends in .git. Without the suffix, it treats the ref as part of the repository name.
  2. Run lola sync to fetch and install the module:

    $ lola sync

    Lola clones the module, extracts skills and commands, and writes them into the target assistant’s native directory structure in the project directory.

lola sync does not refresh a module that it already installed in this project: it skips the Git fetch while the module is registered, and it skips an assistant that already has the installation. New commits in the module repository do not reach the project. To install the current content, run lola mod update <module_name> and then lola update <module_name>. Without a module name, both commands act on every registered module. lola mod update replaces the module copy that every project shares, and lola update rewrites it in every project where it is installed, so save local edits there first.

  1. Optional: To automate the sync at workspace startup, add a postStart command to your devfile:

    commands:
      - id: lola-sync
        exec:
          commandLine: lola sync
          workingDir: /projects/<project_name> (1)
          component: <editor_container> (2)
    events:
      postStart:
        - lola-sync
    1 Replace with the name of your project directory.
    2 Replace with the name of a container component from the components section of your devfile.
Verification
  1. Start a workspace with the .lola-req file in the project root.

  2. Run lola sync (or verify it ran via postStart).

  3. Launch the AI assistant and verify that the skills and commands from the module are available.

Advantages

  • Skills and commands are maintained centrally in Git.

  • Multiple modules can be combined in a single .lola-req file.

  • Supports multiple AI assistants from a single module.

  • Lola translates modules into each assistant’s native file structure with no runtime overhead.

  • Works across clusters and Che instances.

Limitations

  • Lola must be installed in the workspace container image or installed at startup.

  • Manages skills, commands, agents, instructions, and MCP server declarations. Does not persist assistant-specific settings or plugin caches.

  • Changes to skills and commands are not synced back to the source repository. The module repository is the single source of truth.