# SME Agent Drafts

Before the writer touches a cybersecurity topic, your brand's Subject Matter Expert writes the complete technical brief it has to work from - the facts, mechanisms, standards and common mistakes a practitioner would insist the article get right. Every brief the expert has ever written lives here, whether it is still being researched, waiting on you, or already an article.

![SME Agent Drafts](/docs/images/sme_agents_drafts.png)

Cybersecurity brands only, automatic. The SME Agent runs on every new [Article](/docs/using-gracker/content/articles) - News, Blog Post, Listicle, Product Comparison - once your brand is classified as cybersecurity. There's no toggle to find. Content Optimisation runs without it.

---

## What the SME Agent Does

Your brand's expert - name, background, what they cover, what they hand off, the positions they hold, and what they will never claim - reads the prompt on every generate run, researches it on the web, and writes the brief: the mechanism end to end, the exact rules from the governing standard with a source URL, the failure modes, and what to do about them. The writer then turns that brief into the article, adding nothing of substance on its own.

A handful of items in every brief are marked required. The article is checked against them before it reaches you - anything required that's missing, wrong, or unsupported sends the draft back for one automatic rewrite.

---

## Setting Up Your Expert

Go to **Configuration → SME Agent** in the sidebar. Two ways to get one:

| Way | What Happens |
|---|---|
| **Fill with Gracker Agent** | Researches your brand and writes the whole profile for you |
| **By hand** | Type each field in yourself |

The Gracker Agent search runs **once per brand per day** - it's a full research run, not a quick lookup. Ask again sooner and you're told when you can next, with a reminder that you can still edit the profile by hand in the meantime, and every change you make there is used right away.

Re-running it later ("Find a different expert?") replaces **everything on all three tabs**, including anything you've typed and haven't saved yet.

### The Three Tabs

| Tab | Fields |
|---|---|
| **Expert** | Job title (their role, no company name), short bio, topics they cover, topics they skip |
| **Knowledge** | Specialties, strong opinions (with the reason - these become the expert's take in an article), common mistakes they warn about |
| **Rules & sources** | Never claim (a draft that does gets sent back), trusted sources (as "Name - https://url") |

Every list field takes one entry per line. **Save** and **Cancel** appear at the top only once you've changed something; **Cancel** reverts. A job title is required before you can save.

---

## Where to Find Your Drafts

Go to **Content → SME Agent Drafts**. The sidebar badge counts the drafts waiting for your review.

---

## Draft States

| State | Meaning |
|---|---|
| **Researching** | The expert is still writing this brief |
| **Waiting for you** | The brief is ready - open it, edit if needed, and approve |
| **Approved** | You approved it; the article is written from it and nothing else |
| **Failed** | The expert couldn't finish the brief. Hover the badge to see why. Failed drafts don't open |

The list shows **Prompt**, **Expert**, **Updated**, and **Status**.

---

## Reviewing a Draft

Click a draft that's waiting for you to open the review screen: the question it answers as the heading, **Written by** the expert's name - click it to read the full expert profile - and the brief itself, editable in a **Visual** or **Markdown** view. Below it, an expandable list of every page the expert read while researching.

| Action | What It Does |
|---|---|
| **Save changes** | Keep your edits without approving yet |
| **Regenerate** | Throw this brief away and have the expert write a new one from scratch |
| **Approve and write the article** | Lock in the brief (with any edits) and start writing |

The article is written from this brief and nothing else. Fix anything wrong, add anything missing, before you approve - not after.

---

## After You Approve

The article is generated from the approved brief. Once it's ready, open it straight from the drafts list or with **Open the article** on the review screen. Editing the brand's expert afterwards changes future drafts, not one already approved.

---

## FAQ

**Can I skip the SME Agent for one piece?**

  Not for Create New Content - it runs automatically on every generate run for a cybersecurity brand, with no per-piece toggle. Content Optimisation always runs without it.

**What happens if I click Regenerate?**

  This brief is thrown away, including any edits you made, and the expert writes a new one from scratch. The article still can't be written until you approve whatever comes back.

**Can I search or filter the drafts list?**

  Yes. Search with **Search questions and experts…**, and filter by state: All drafts, Researching, Waiting for you, Approved, or Failed.

**Why can't I open a researching draft?**

  There's nothing to read yet - the expert hasn't finished writing it.

**Can I have more than one expert per brand?**

  No - one expert per brand, used on every generate run.

**Do I need to fill in every field on the expert?**

  Only the job title is required to save. The more you fill in, especially trusted sources and never claim, the more the brief - and the article - actually reflects your brand's expertise.

---

## Next Steps

- [Write an article](/docs/using-gracker/content/articles): generate the piece the expert's brief feeds into
- [Cybersecurity Standards](/docs/advanced/industry-specialised/cybersecurity-standards): how NIST, MITRE and CISA references get added on top

- [Trust Score](/docs/using-gracker/aeo-audit/trust-score): how a published page's security claims get scored
- [Find your content gaps](/docs/advanced/content-gaps): competitor pages AI cites instead of yours
