Manage Claude Code skills for your team
Approve a skill once and it lands in Claude Code on every developer's machine, at the approved version. Ban one and it's removed everywhere, and blocked the moment anyone invokes it. Every run is recorded in a tamper-evident audit log.
- Skills, every project
- ~/.claude/skills
- Skills, in a repository
- .claude/skills
- Also installed
- Subagents, slash commands
- Blocks at invocation
- Yes, through its hook
Claude Code is the one agent with a hook that can refuse a tool call, so skillsvault blocks a banned skill there the moment it's invoked, and the developer sees which rule stopped it. The skillsvault plugin brings the skill, the MCP server and the hook in one install.
Set it up
Paste this into Claude Code. It reads the setup guide, installs what it needs and asks you to sign in once:
Or by hand, from a terminal:
Then skillsvault install-hook turns on the gate, or install the Claude Code plugin, which includes it. See the runtime gate.
Questions
- Where does Claude Code look for skills?
- For every project: ~/.claude/skills. Inside a repository: .claude/skills.
- Can skillsvault block a skill in Claude Code?
- Yes. Claude Code has a hook that can refuse a tool call, so the skillsvault gate stops a banned skill when it's invoked and tells the developer which policy stopped it and why.
- What does skillsvault install into Claude Code?
- Your organisation's approved skills, subagents (~/.claude/agents) and slash commands (~/.claude/commands), each at the version the registry approves.
- Is it free?
- Free for one developer. Teams pay per member: €20 a month on Team, €40 on Business.
One registry for all of them
The same approved list reaches every agent your team uses.
Start with the skills you already have
skillsvault scan lists every skill on this machine across all your agents and flags the risky ones. It runs locally, with no account.