Skip to content

SkillKeeper

SkillKeeper installs and manages skills and hooks for AI coding agents, both per-project and globally on a machine.

A skill is a directory containing a SKILL.md file plus supporting files. A hook is an optional privileged add-on inside a skill that modifies an agent's own configuration. Skills are distributed through Git repositories (including private GitHub and Bitbucket repositories) and are installed into the locations expected by each supported agent.

Supported agents

Claude, Codex, Copilot, Cursor, OpenCode.

Platforms

Linux, macOS, Windows.

Two front ends, one core

SkillKeeper ships two front ends over one shared, framework-agnostic core:

  • CLI - the primary interface for v1; suitable for scripting and headless use.
  • Desktop app - Tauri v2 (Rust backend) + a React 19 renderer. The renderer reaches the backend through a typed Tauri command/event bridge; the backend is the authority for the filesystem, Git, config, and state.
  • Getting Started - install, add a repository, install your first skill.
  • Skills and Hooks - skill format, hook strategies, resolution schemes.
  • Repositories - adding repositories, branch tracking, update detection, and how Git is invoked.
  • Projects - tracked project folders, per-project agent selection, and reconciliation.
  • MCP Servers - MCP presets, the mcp.yml format, install/update/ remove, and per-agent native config.
  • CLI Reference - every command and option.
  • Configuration - config.yaml sections and defaults.
  • Architecture - package graph, process boundaries, domain model.

What SkillKeeper does NOT do (v1)

  • Vulnerability scanning (a seam exists; no implementation in v1).
  • Publishing skills (SkillKeeper consumes repositories; it does not author them).
  • Auto-resolving SSH key passphrases (delegated to the user's ssh-agent).
  • Detailed desktop UI screens beyond the shell.