> ## Documentation Index
> Fetch the complete documentation index at: https://hashbot.com/docs/llms.txt
> Use this file to discover all available pages before exploring further.

# Cleanse

> Once you've configured your **Manager Roles**, **Allowlisted Roles & Users**, and **Name-Filters**, it's time to retroactively scan and clean up your server.

The `/cleanse` command instructs Hashbot to scan all existing members against your current filters — including users who joined **before Hashbot was installed** — and sends the results to the [Dashboard](https://dashboard.hashbot.com) for review.

<Tip>
  After running `/cleanse`, head to the [Hashbot Dashboard](https://dashboard.hashbot.com) to review flagged users and take action — it is the recommended way to manage cleanse results.
</Tip>

### What Is Cleanse?

Cleanse is a **manual, retroactive scan**.
Unlike Hashbot's real-time protection (which acts on new joins or name changes), `/cleanse` reviews your **entire member list** to catch **previously undetected threats**.

### How to Use `/cleanse`

1. Run `/cleanse` in any channel where Hashbot is active.
2. Hashbot scans your server against all active name filters. This may take a few seconds or minutes, depending on server size.
3. You'll receive a **summary embed** showing the total number of matches found.
4. Click the **"Review on Dashboard"** button in the embed to open the full detection report on the [Hashbot Dashboard](https://dashboard.hashbot.com).
5. On the Dashboard, review each flagged user and take action — **Ban**, **Mark Safe**, or **Skip**.

### What If I See a False Positive?

If someone is flagged by mistake:

1. Click **Mark Safe** on the Dashboard to dismiss the match.
2. Add them to the allowlist so they are not flagged again:

`/allowlist add @user`
Or by ID:
`/allowlist add-id 123456789`

<Info>This ensures trusted users are exempt from filters, even retroactively.</Info>

### Best Practices

* Use `/name-filters test` before running a cleanse to verify your filters
* Review your filter list for accuracy before scanning
* Pause actions with `/settings pause pause-state:pause` if needed
* Use the Dashboard to work through flagged users at your own pace — results persist until reviewed