Skillbase / spm

Installation & Setup

Install spm and load your first AI skill in under a minute.

Install spm

Install the Skills Package Manager globally via npm:

npm install -g @skillbase/spm

Verify the installation:

spm --version

Initialize

Create the global skills directory:

spm init

This creates ~/.spm/ with the following structure:

~/.spm/
├── installed/     # Installed skills live here
├── index.json     # Compact skill index (auto-generated)
├── feedback.json  # Usage feedback data
└── config.json    # Configuration

For project-scoped skills, use --project to initialize in the current directory:

spm init --project

This creates a .spm/ directory and a skillbase.json manifest in the project root for declaring dependencies.

Install your first skill

spm add core/docx

This downloads the skill, validates it, resolves any dependencies, and updates the index.

Check it's installed:

spm list
spm info core/docx

Connect to an AI client

spm works with any MCP-compatible AI client. Connect with a single command:

spm connect claude

Supported clients:

ClientCommand
Claude Desktopspm connect claude
Zedspm connect zed

This updates the client's MCP configuration to include spm as a server. The AI model can now discover and load skills automatically.

Manual MCP configuration

If your client isn't directly supported, add spm to the MCP server configuration manually:

{
  "mcpServers": {
    "skills": {
      "command": "spm",
      "args": ["serve", "--stdio"]
    }
  }
}

How it works

Once connected, the AI model:

  1. Receives a compact index of all installed skills via skill_list
  2. When a task matches a skill's trigger, calls skill_load to pull in the full instructions
  3. Executes the task using the loaded skill
  4. Reports feedback via skill_feedback to improve future recommendations

The model does all of this automatically — you just ask it to do something and it picks the right skill.

Next steps

NextCreating Skills