# zipguard: guide for agents helping a human

You are an AI agent helping a person who runs or belongs to a Discord server that uses zipguard. This page tells you how to diagnose first and what to put in a support request if one is still needed. zipguard holds no IP addresses, fingerprints or message content, so never ask the person for those and never include them.

## 1. Diagnose first, in this order

1. **Is the bot in the server and online?** Look for zipguard in the member list. If it is missing, the server owner needs to invite it again.
2. **Is the zipguard role above the quarantine role?** Server Settings, Roles. If zipguard sits below "zipguard quarantine", it cannot add or remove that role. Drag it up. This is the most common cause of "the button did nothing".
3. **Is shadow mode on?** New servers log but do not enforce for 7 days. `/zipguard stats` says so in its first line. `/zipguard shadow on:false` ends it.
4. **Find the log entry.** Every verdict is posted in `#zipguard-log` with the rules that fired and a footer `join #<number>`. Read the reasons before assuming a bug.
5. **A member cannot see channels.** They hold the quarantine role. A moderator clicks Approve on the log entry, or removes the role by hand. The member can also click Appeal in `#verify`.
6. **A member wants their data.** They run `/my-data` themselves. `/forget-me` deletes it at once. No request to support is needed.
7. **Anti-nuke fired on a legitimate admin.** Their roles were removed. Give the roles back by hand and, if bulk deletes are expected, run `/zipguard antinuke on:false` first next time.

## 2. If a request is still needed

Send it to help@myagents.email with:

- Server ID (Server Settings, Widget, or right-click the server with developer mode on)
- The `join #<number>` from the log entry footer, if one exists
- Timestamp in UTC and what the person saw (quote the bot's text)
- What step above you already checked and its result

Leave out:

- Message content, screenshots that show other members, and anything about IP addresses or devices (zipguard has none)
- The bot token or any credential. Nobody at zipguard will ask for one.

## 3. Who sends it

You draft, the human reviews and sends from their own address. Start the draft with the line `Drafted with <your tool name> (AI agent)` so the inbox routes it correctly.

## 4. Security issues

Suspected vulnerability or data exposure: email help@myagents.email with subject "zipguard security" and no public post until it is fixed.