> ## Documentation Index
> Fetch the complete documentation index at: https://docs.turen.io/llms.txt
> Use this file to discover all available pages before exploring further.
# Skill Delivery
> Push custom Claude Code skills to your entire team
Skill delivery lets you create custom Claude Code skills and automatically distribute them to every developer machine in your organization. Skills appear as native slash commands in Claude Code: developers can invoke them manually or Claude can invoke them automatically based on context.
## What Are Skills?
Skills are Claude Code extensions that define workflows, guidelines, and tools. They're written as Markdown files with YAML frontmatter and can include:
* **Code review checklists** that Claude follows when reviewing PRs
* **Deployment procedures** with step-by-step instructions
* **API design guidelines** that Claude applies when writing endpoints
* **Security audit workflows** that trigger automatically when discussing vulnerabilities
* Anything else you want Claude to know or do consistently across your team
## How It Works
Upload a skill package (`.tar.gz`) in the dashboard under **DevEx > Skill Delivery** (`/devex/skills`).
The skill is encrypted with your organization's key and stored securely in S3.
Within 15 minutes, every Turen agent downloads and deploys the skill to `~/.claude/skills/turen/`.
Claude Code discovers the skill natively. Developers see it in their `/` command menu and Claude can invoke it automatically.
## Creating a Skill
A skill is a `.tar.gz` archive containing a `SKILL.md` file with YAML frontmatter:
```markdown theme={null}
---
name: security-review
description: >
Security-focused code review guidelines. Use when reviewing
code, auditing security, or when asked about vulnerabilities.
user-invocable: true
allowed-tools: Read, Grep, Glob
---
## Security Review Checklist
When reviewing code for security issues, always check for:
### Input Validation
- [ ] All user input is validated and sanitized
- [ ] SQL queries use parameterized statements
- [ ] File paths are validated against traversal attacks
### Authentication
- [ ] Passwords are hashed with bcrypt/argon2
- [ ] Session tokens have proper expiry
- [ ] API keys are not hardcoded
### Data Protection
- [ ] Sensitive data is encrypted at rest
- [ ] PII is not logged
- [ ] Secrets are loaded from environment variables
```
### Frontmatter Options
| Field | Required | Description |
| -------------------------- | -------- | ---------------------------------------------------------------------- |
| `name` | Yes | Skill identifier: becomes the `/slash-command` name |
| `description` | Yes | When to use the skill. Claude uses this to decide when to auto-invoke. |
| `user-invocable` | No | If `true` (default), appears in the `/` command menu |
| `disable-model-invocation` | No | If `true`, only the developer can invoke it: Claude won't auto-invoke |
| `allowed-tools` | No | Restrict which tools Claude can use (e.g., `Read, Grep, Glob`) |
### Supporting Files
Skills can include additional files alongside `SKILL.md`. For example, a deployment skill might include a checklist:
```text theme={null}
deployment-guide/
├── SKILL.md
└── checklist.md
```
The `SKILL.md` can reference these files and Claude will have access to them.
## Uploading a Skill
Create a `.tar.gz` archive containing your `SKILL.md` and any supporting files.
```bash theme={null}
tar -czf security-review.tar.gz SKILL.md
```
In the dashboard, navigate to **DevEx > Skill Delivery** (`/devex/skills`). Select the **Custom Skills** tab.
Click **Upload Skill** and select your `.tar.gz` file (max 10MB). You can also drag and drop the file. The dashboard validates the package and extracts the skill metadata.
The skill is now queued for delivery. Agents will pick it up within 15 minutes.
## Managing Skills
The Skill Delivery page has two tabs: **Custom Skills** and **Curated Skills**.
### Custom Skills
The Custom Skills tab shows a table of all skills you've uploaded. From here you can:
* **Enable/disable**: Toggle whether a skill is delivered to agents
* **Edit**: Update the skill name, description, or upload new content
* **Download**: Download the current skill package
* **Version history**: View all previous versions with version number, content hash, file size, and timestamp
* **Restore**: Roll back to any previous version from the version history
* **Delete**: Remove a skill from all agent machines
Use the filter bar to search skills by name or status (enabled/disabled).
### Curated Skills
The Curated Skills tab shows pre-built skills created by the Turen team. These cover common use cases and can be enabled or disabled with a single toggle. You cannot edit or delete curated skills.
## Versioning
Every time you upload or edit a skill, a new version is created. The version history shows:
* Version number
* Content hash
* File size
* Who made the change and when
* A **Restore** button to roll back to any previous version
## What Developers See
Once a skill is deployed, developers see it natively in Claude Code:
**Manual invocation:**
```text theme={null}
> /security-review src/auth.js
```
**Auto-invocation**: if the skill's description matches the conversation context, Claude invokes it automatically:
```text theme={null}
Developer: "Review this code for vulnerabilities"
Claude: [automatically loads security-review skill]
```
Skills appear seamlessly alongside Claude Code's built-in commands. Developers don't need to install or configure anything.
## Use Cases
Define what Claude should check during code reviews: security, performance, style, accessibility.
Step-by-step checklists that Claude follows when helping with deployments.
Enforce consistent API patterns, naming conventions, and error handling across your team.
Help new developers navigate the codebase with context Claude can reference.
## Security
* Skills are encrypted with your organization's unique key (AES-256-GCM) before storage
* Only agents authenticated with your organization's credentials can download skills
* Skills are written to a dedicated `~/.claude/skills/turen/` directory: they cannot overwrite developer-created skills
* All skill changes are logged in the audit trail