SkillzSAFE
An MCP server for loading skills (shim for non-claude clients).
Overview
From the repository's own README, as read at the audited commit. Badges and raw HTML are left out.
👌 Use _skills_ in any agent (Codex, Copilot, Cursor, etc...)
[](https://pypi.org/project/skillz/) [](https://pypi.org/project/skillz/)
⚠️ Experimental proof‐of‐concept. Potentially unsafe. Treat skills like untrusted code and run in sandboxes/containers. Use at your own risk.
Skillz is an MCP server that turns Claude-style skills (`SKILL.md` plus optional resources) into callable tools for any MCP client. It discovers each skill, exposes the authored instructions and resources, and can run bundled helper scripts.
💡 You can find skills to install at the [Skills Supermarket](http://skills.intellectronica.net/) directory.
Quick Start
To run the MCP server in your agent, use the following config (or equivalent):
{
"skillz": {
"command": "uvx",
"args": ["skillz@latest"]
}
}with the skills residing at ~/.skillz
or
{
"skillz": {
"command": "uvx",
"args": ["skillz@latest", "/path/to/skills/direcotry"]
}
}or Docker
You can run Skillz using Docker for isolation. The image is available on Docker Hub at intellectronica/skillz.
To run the Skillz MCP server with your skills directory mounted using Docker, configure your agent as follows:
Replace /path/to/skills with the path to your actual skills directory. Any arguments after intellectronica/skillz in the array are passed directly to the Skillz CLI.
{
"skillz": {
"command": "docker",
"args": [
"run",
"-i",
"--rm",
"-v",
"/path/to/skills:/skillz",
"intellectronica/skillz",
"/skillz"
]
}
}Gemini CLI Extension
A Gemini CLI extension is available at intellectronica/gemini-cli-skillz.
Install it with:
gemini extensions install
f6e16c6f32edOBSERVED · 2026-10-01Connect
Built from this server's own package name, version and transport as found in its source — not copied from anyone's documentation, so it cannot drift against a page we do not control.
claude mcp add skillz -- uvx skillz
{
"mcpServers": {
"skillz": {
"command": "uvx",
"args": [
"skillz"
]
}
}
}Exposed tools (1)
1 read · 0 write · 0 destructive.
| Tool | Risk | Description |
|---|---|---|
_skill_tool | read | LOGGER.info( |
Trust audit
SAFEgrade B · trust 89/100 Nothing in the source contradicts what it says it does. Grade A is reserved for packages that have also passed the behavioural sandbox.
| Layer | What it checks | Result |
|---|---|---|
| L0 | Provenance & inventory | PASS |
| L1 | Static analysis of the code | PASS |
| L2 | Instruction surface (what it tells the agent) | PASS |
| L3 | Class-specific surface | PASS |
| L4 | Behavioural (sandbox) | SKIPPED |
What the source does
- Filesystem
- none-observed
- Network
- none-observed
- Shell
- none-observed
- Dependencies
- pinned
- Secrets in source
- none-found
Findings (3)
resource_uri="resource://skillz/testskill/../../../etc/passwd"
decoded = base64.b64decode(result["content"])
decoded = base64.b64decode(result["content"])
Gates applied: no_behavioural_pass.
f6e16c6f32edfull audit observations/trust-audit/mcp-server/intellectronica__skillz-1.json · Report an issue / request a re-scanAudit history
Every audit this server has had. A grade with a past is a grade somebody is still checking.
| Date | Source | Verdict | Grade | Score | Change |
|---|---|---|---|---|---|
| 2026-10-01 | f6e16c6f32ed | SAFE | B | 89 | first audit |
Questions
What is the Skillz MCP server?
An MCP server for loading skills (shim for non-claude clients).
What tools does Skillz expose?
1 in total: 1 read-only, 0 that write, and 0 that can delete or overwrite. Every one is listed on this page with its risk.
Is Skillz safe to connect to an agent?
The audit found nothing in the source that contradicts what it says it does, and graded it B (89/100). Grade A is held back for packages that have also passed a sandboxed behavioural run, which is why a clean server reads B.
What credentials does Skillz need?
No credential environment variables were found in its source, so it appears to need none.
How current is this page?
The grade is for one exact copy of the source (f6e16c6f32ed), read on 2026-10-01. The repository is watched and re-audited when it changes.