Are you the author? Sign in to claim
SkillKeeper installs and manages skills and hooks for AI coding agents (Claude, Codex, Copilot, Cursor, OpenCode). Skill
SkillKeeper installs and manages skills and hooks for AI coding agents (Claude, Codex, Copilot, Cursor, OpenCode).
SkillKeeper comes in two forms over one shared Rust core:
skillkeeper) that does the same from your shell
and in scripts.Skills are distributed as Git repositories; a repository can also ship MCP server presets. SkillKeeper clones a repository, resolves what it provides, installs the parts you choose into the agents you target, and tracks everything for updates. Supported agents: Claude, Codex, Copilot, Cursor, and OpenCode.
Desktop app -- download the bundle for your platform from the latest release.
CLI (skillkeeper) -- one line, using only tools already on your system. The
script detects your platform, downloads the matching CLI archive from the latest
release, and adds the binary to your PATH.
macOS / Linux:
curl -fsSL https://raw.githubusercontent.com/lorem-dev/skillkeeper/main/scripts/install.sh | sh
Windows (PowerShell):
irm https://raw.githubusercontent.com/lorem-dev/skillkeeper/main/scripts/install.ps1 | iex
See Getting Started for other options.
Hitting an install problem (for example macOS reporting the app as "damaged")? See Troubleshooting.
The CLI binary is skillkeeper. Every command exits non-zero on failure.
Add a repository of skills, then see what it provides:
skillkeeper repo add git@github.com:example/skills.git
skillkeeper skill list
Install a skill for an agent (per-project by default; add --global for
machine-wide). Hooks are privileged and only installed with --allow-hooks:
skillkeeper skill install <skill-id> --agent claude
skillkeeper skill install <skill-id> --agent codex --global --allow-hooks
Track a project, then check for available updates across its repositories and installed skills:
skillkeeper project add .
skillkeeper check
Inspect, verify against the manifest, and repair an installation:
skillkeeper skill info <skill-id>
skillkeeper skill verify <skill-id>
skillkeeper skill repair <skill-id>
Supported agents are claude, codex, copilot, cursor, and opencode.
Run skillkeeper --help (or <command> --help) for the full command set, also
documented in the
CLI Reference.
A repository is any Git remote (SSH or HTTPS, public or private) that contains
one or more skills, and optionally MCP server presets. A skill is a directory
with a manifest plus the files it installs; hooks are opt-in edits a skill can
make to an agent's configuration. To publish your own skills, create a Git
repository with that layout and point SkillKeeper at it with repo add.
See Repositories and Skills and Hooks for the repository format and authoring guide.
skillkeeper-test-repo is a
fixture repository that exercises every documented resolution path: flat and
grouped skills, both hook layouts, guidance-file precedence, MCP presets at the
repository root and inside a group, and a deliberately unresolvable skill. It is
wired into this repository as a Git submodule at examples/test-repo. Fetch it
with git submodule update --init (the submodule uses an SSH remote, so this
needs a GitHub SSH key; the fixture is not required by any build or test).
Entry points -- the links go to GitHub, the paths are where the files land once the submodule is checked out:
minimal-skill/SKILL.md
-- the smallest valid skill manifest.json-hooks-skill/hooks/pre-tool-use/HOOK.md
-- a hook that merges into an agent's JSON config.mcp.yml
and
tooling/mcp.yml
-- root and group-scoped MCP presets.Links into a submodule cannot be relative: this repository's tree stores
examples/test-repo as a commit pointer, not as files, so a relative path would
404 on GitHub.
SkillKeeper is a Rust + pnpm monorepo: a Cargo workspace of domain crates
(skillkeeper-core, skillkeeper-config, skillkeeper-agents,
skillkeeper-cli) with a Tauri v2 + React desktop app, all over one shared
Rust core.
See the Development guide for setup, commands, and conventions.
Apache-2.0. See LICENSE.
Hooking implementations and supporting tools for various coding agents (Claude, Cursor, Gemini, etc)
Claude Code hook that writes a forward-only why-block (decisions, trade-offs, assumptions, limitations) into your PR des
Blocks dangerous git and shell commands from being executed by AI coding agents
One command to install 6 essential safety hooks in 10 seconds — zero dependencies