User guide

# IZNIR — User guide

**IZNIR** is the **Certified Skill Factory**: it turns a description of what you want an AI to do into a structured, tested, certified skill you can export and deploy.

This guide is for **people using the web app**, not for server installation (see the project **README** for developers).

---

## What you can do

- **Create skills** in a guided six-step flow: from your idea → definition → guardrails → validation → automated tests → certification.
- **Keep skills private** to your workspace, or use the **marketplace** to discover and fork published skills (when your deployment enables it).
- **Export** certified skills as JSON, YAML, or formats aimed at common AI platforms.
- **Manage privacy**: request a copy of your data or account deletion where your administrator has enabled self-service tools.

---

## Signing in and workspace

1. **Register** to create your first workspace (you become **Owner**).
2. **Sign in** with your email and password.
3. Everything you build lives in **your workspace** unless you publish to the marketplace.

**Roles (short):**

| Role        | Typical use                                      |
| ----------- | ------------------------------------------------ |
| Member      | Create and edit skills, use marketplace.       |
| Owner       | Invite members, view workspace activity.       |
| Superadmin  | Platform administration (rare; special users). |

---

## Creating a skill (the six steps)

### 1. Intent

Write a clear description: domain (e.g. legal, finance), what inputs you expect, and what outputs you need. The more precise you are, the better the next steps.

### 2. Definition

Review the generated **name**, **description**, **input/output shape**, **scope** (allowed behavior), and **out of scope** (must refuse). Edit anything that is wrong before continuing.

### 3. Guardrails (review)

The system proposes **guardrails**: safety rules such as handling uncertainty, staying in scope, and audit expectations. Read them carefully; they define how the skill is allowed to behave.

### 4. Validate

You must **confirm each guardrail** before you can test. This is a deliberate gate: the server enforces it, not just the screen.

### 5. Test

Run the **test suite**. You need **at least 80%** to move on. If you fail, read the failing cases, improve your definition or guardrails, and run tests again.

### 6. Certify

When tests pass, **certify** the skill. You will get certification metadata and can **export** the artifact for your pipelines.

---

## Versions

Creating a **new version** starts a **new draft**. That draft must go through validation and testing again before it can be certified. Older certified versions remain for traceability.

---

## Marketplace

- **Browse** listings, ratings, and descriptions.
- **Fork** a listing into your workspace to adapt it under your own review process.
- Always verify **publisher** and **version** before using a skill in production.

---

## Privacy and your rights

- Use **Settings → Privacy** for export, deletion requests, and consent where available.
- The public **Privacy & GDPR** page explains how data is handled at the service level for your deployment.
- Never put **passwords or API keys** into skill text or contact forms.

---

## Getting help

- **Help center** (`/help`) — workflows and FAQ-style topics.
- **Glossary** (`/glossary`) — definitions of terms used in the product.
- **Contact** (`/contact`) — message the team running your instance.

---

*Document version aligns with app releases; see **Changelog** for what changed.*