Admin quickstart
From upload to your people using it, in about ten minutes.
You need the Policies: edit permission to upload and decide, and Settings: manage to connect a repository.
1. Upload
- Open Govern → Skills & Plugins.
- Click Upload and pick a
SKILL.md, a zipped skill folder, or a zipped Claude Code plugin.
Waxell reads it, works out whether it's a skill or a plugin, and scans it. The drawer opens on the new submission; the scan result appears within a few seconds.
2. Review
The drawer shows everything you need to decide:
| Section | What it tells you |
|---|---|
| Security scan | The verdict (Safe, Caution, Do not install), a 0–100 risk score, the worst finding, and each finding with the file and line it came from. |
| What it contains | The skills, commands and agents it adds — and, flagged, any hooks, MCP servers or scripts: code that runs on its own. |
| Files | Every file in the upload, readable in place. |
The scanner's overall verdict can read Safe while it reports a high finding. Waxell shows the worst finding beside the verdict — and your policy acts on the finding, not just the verdict.
3. Decide
At the bottom of the drawer:
| Button | Does |
|---|---|
| Approve | Publishes it. Choose Install mode first: Available — people install it, or Required — enabled for everyone. |
| Reject | Declines it. Add a Note — the requester sees it. |
| Rescan | Scans again (for example after the scanner was updated). |
| Revoke | (On an approved item) Takes it back out of everywhere it was published. |
| Approve anyway | (On a rejected or revoked item) Overrides the earlier decision. |
Approving a new version of something already approved revokes the old one automatically — only one version is ever published.
4. Publish
Approved items go to two places. Set up whichever your people use:
- Claude Code — your approved marketplace. See Claude Code.
- Claude Desktop, Cowork and claude.ai — a GitHub or GitLab repository that claude.ai syncs. See Sync to claude.ai.
5. Automate the easy ones (optional)
Turn on Auto-approve clean scans under Settings so a clean skill doesn't wait for a person. See Auto-approval.
The queue
| Tab | Shows |
|---|---|
| Needs review | Waiting on a person, or still being scanned. |
| Approved | Published. |
| Rejected | Rejected or revoked. |
| All | Everything. |
Search by name, description or who submitted it. Every decision — yours or the policy's — is recorded in your workspace's audit log, with who decided and why.